MCP App Store
by KeenEthics
Specter Systems icon

Specter Systems

By Specter Systems

Overview

Specter Systems helps merchants and brokers complete business financing intake, upload bank statements and supporting documents, and track financing applications. Authenticated Specter users can review their own recent applications, facilities, lender submissions, and offer status.

Tools

accept_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Approves a pending Heartbeat action so it can be applied. Confirm with the user first. Optionally set the customer lifecycle. After accepting, call execute_heartbeat_action to apply it. May dispatch a prepared message to external recipients. Operates on one stored proposal by ID; it does not accept arbitrary operation names, register tools, or expose hidden model-callable tools.

add_note

ChatGPT
For non-merchant Specter broker/operator users only. Adds a note to a CRM record's activity timeline (lead, merchant, deal, or submission). Use when the user asks to add a note or log a comment on a record. 'entity_type', 'entity_id', and 'body' are required. This writes live CRM/FOS data.

archive_fos_setup_deal_status

ChatGPT
Tenant-admin-only setup helper. Archives a deal/facility status definition by stable value in the authenticated tenant. Use only after explicit admin confirmation.

archive_fos_setup_group

ChatGPT
Tenant-admin-only setup helper. Deletes/archives a manual group for Deals, Leads, or Merchants by ID, scoped to the authenticated tenant and entity type. Use only after explicit admin confirmation.

bulk_update_facilities

ChatGPT
Applies ONE uniform update to up to 100 facilities/applications in a single call — use this instead of many update_facility calls when the user asks to move a set of deals to the same status (e.g. "move these 9 to closed_follow_up"). Every facility gets the same field values; for per-deal tailored notes, follow with update_facility per record. Returns a per-ID success/failure list. Backend tenant isolation, permission checks, and audit history apply to each record.

close_deal

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL deal close-out on a facility: transitions the lifecycle stage, optionally sets the tenant deal status slug, creates a follow-up task, and adds a note — with per-step results. PREFER THIS over separate transition_application_stage + update_facility + create_task + add_note calls when the user asks to mark a deal funded/closed and schedule follow-up. Steps run in order and stop on the first failure. Backend lifecycle gates, tenant isolation, permission checks, and audit history apply.

compare_application_offers

ChatGPT
Reads and compares available offers for one authenticated facility/application. Use when the user asks which offer is better or wants financing terms side by side. This is read-only.

convert_lead_to_facility

ChatGPT
For non-merchant Specter broker/operator users only. Converts an existing lead into a deal/facility (auto-creates the merchant if missing and advances the lead). Use when the user asks to convert a lead to a deal. Provide lead_id; product_type and requested_amount are optional. This writes live CRM/FOS data.

create_application_facility

ChatGPT
Creates an authenticated facility/application record inside the user's Specter tenant. Use for internal users or brokers who explicitly ask to create a facility after authentication. This is additive but writes live CRM/FOS data.

create_capital_source

ChatGPT
For non-merchant Specter broker/operator users only. Merchant applicant accounts must not use this tool. Creates a new lender-board/capital-source record in the authenticated user's Specter tenant after the user clearly asks to add the lender. Use for lender details extracted from chat, files, PDFs, or product guides. Do not pass tenant_id, user_id, API keys, portal credentials, bank account data, tax IDs, SSNs, or other secrets. Backend tenant scope, RBAC, duplicate-name checks, validation, and audit events remain authoritative.

create_communication_channel

ChatGPT
Creates the communication channel for an entity that does not have one yet. Use ONLY after get_communication_channel_for_entity returns channel: null — a lead or counterparty nobody has messaged has no channel, and send_channel_message requires a channel_id it cannot invent. Provide entity_type, entity_id, and a display_name (the business or person's name, which is what the thread is labelled with). Creating the channel does not message anyone: it opens the thread so a later send has somewhere to land. Tenant-scoped and RBAC-enforced; if a channel already exists, read it with get_communication_channel_for_entity instead of creating a second one.

create_fos_setup_deal_status

ChatGPT
Tenant-admin-only setup helper. Creates a deal/facility status definition in the authenticated tenant only. Use for admin setup assistance when a required deal status is missing.

create_fos_setup_group

ChatGPT
Tenant-admin-only setup helper. Creates a manual group for Deals, Leads, or Merchants in the authenticated tenant only. Use for admin setup assistance when a required board/list group is missing. Never invent cross-tenant IDs and do not use inside Workflow Builder automation design.

create_lead

ChatGPT
For non-merchant Specter broker/operator users only. Creates a new lead (prospect) in the authenticated user's Specter tenant. Use when the user explicitly asks to add or create a lead. Only 'name' is required; include contact and business details when the user provides them. This writes live CRM/FOS data.

create_lender_submission

ChatGPT
Creates an authenticated lender submission for a specific facility/application. Use only when the user explicitly asks to submit or prepare a facility for a lender. This writes live submission pipeline data.

create_merchant

ChatGPT
For non-merchant Specter broker/operator users only. Creates a new merchant / counterparty (business) in the authenticated user's Specter tenant. Use when the user asks to add a merchant or business. 'name' and 'type' are required; default 'type' to "business" for a company. Optional business contact and profile fields are only for populating the CRM record when the user provides them. Do not request tax IDs or other government identifiers in chat; enter those through the secure Specter portal. This writes a business record, not a login account, and does not grant merchant access.

create_message_draft

ChatGPT
Writes a proposed reply to a conversation and stores it for a human to review. It does NOT send anything: the draft waits until a person approves it. Use this whenever you want to suggest a response — for example after an inbound reply to an outreach message. Read the thread first with list_conversation_messages so the draft answers what was actually said. A conversation can hold only one pending draft; if one already exists this returns a conflict.

create_offer

ChatGPT
For non-merchant Specter broker/operator users only. Creates a financing offer, optionally linked to a facility/deal or submission. Use when the user asks to add or record an offer. 'funded_amount' is required; link it with facility_id when known. This writes live CRM/FOS data.

create_task

ChatGPT
For non-merchant Specter broker/operator users only. Creates a task attached to a CRM record (lead, merchant, deal, or submission). Use when the user asks to add a task, follow-up, or to-do. 'entity_type', 'entity_id', 'task_type', and 'title' are required; use task_type "custom" for a general task. This writes live CRM/FOS data.

create_workflow_draft

ChatGPT
Creates a new Workflow V3 draft for the authenticated tenant. Use only after the user asks to create a workflow and the graph has been validated or intentionally saved as an incomplete draft. This operation only persists workflow configuration; it does not dispatch the configured node actions or register new model-callable tools. The MCP server forces is_draft=true; publish_workflow is required before the draft runs live automation.

decline_and_notify

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL decline: transitions the facility to outcome, optionally sets the tenant declined status slug, sends a merchant-facing decline message (skipped when no message body is given), and records an internal note with the decline reason — with per-step results. PREFER THIS over separate transition + update_facility + send_channel_message + add_note calls when the user asks to decline or kill a deal. Steps stop on the first failure.

defer_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Defers a pending Heartbeat action to revisit later, optionally until a specific time. Use when the user wants to snooze the proposed next step.

delete_accounting_access_accountId

ChatGPT
Revoke a seat's accounting grant Takes the brokerage's books back from one member. They keep their own deals, stakes and payout method. Answers 204; a seat that held no grant is a not-found. The admin seat. DELETE /v1/accounting/access/:accountId (access: workspace:admin)

delete_accounting_allocations_id

ChatGPT
Undo an allocation Reverses a reconciliation: unpicks one allocation, so the deal's collected total falls and every stake's available balance falls with it. Answers 204. This is how a line allocated to the wrong deal or under the wrong type is corrected — unpick it, then allocate it again. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. DELETE /v1/accounting/allocations/:id (access: workspace:books)

delete_accounting_categories_identifier

ChatGPT
Delete an unused transaction category Delete a category that no bank line is filed under (its uses is 0). A category in use is refused with 409 — archive it instead by saving it with active=false, which keeps those lines' history readable. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. DELETE /v1/accounting/categories/:identifier (access: workspace:books)

delete_accounting_deals_id

ChatGPT
Remove a deal from the ledger Destroys a funded deal's economics row, and with it the record of what that deal earned. The admin seat. A deal that has already paid somebody cannot be deleted at all — the payout pins its stake and the stake pins the deal — which answers 409. To take a deal out of the books without destroying it, patch its status instead. DELETE /v1/accounting/deals/:id (access: workspace:admin)

delete_accounting_invoices_id

ChatGPT
Discard a draft invoice Deletes a draft. An issued invoice is never deleted (409): void it with post.accounting.invoices.id.void. Takes the books to KEEP; anyone else is refused with 403. DELETE /v1/accounting/invoices/:id (access: workspace:books)

delete_accounting_me_payout-methods_id

ChatGPT
Remove a payout method Deletes one of the CALLER's own payout methods, and the sealed details with it. A request that already named it keeps standing with no destination. When the default is removed the newest remaining method takes its place. Answers 204. DELETE /v1/accounting/me/payout-methods/:id (access: workspace)

delete_accounting_plaid-connections_itemId

ChatGPT
Unlink a company bank Revokes the linked bank at Plaid, stops its future syncs and removes cached bank data. Imported accounting transactions remain. Use only when the workspace administrator asks to disconnect the bank. DELETE /v1/accounting/plaid-connections/:itemId (access: workspace:admin)

delete_accounting_settlement_statements_id

ChatGPT
Discard a settlement draft Deletes a draft statement. An approved statement is never deleted. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. DELETE /v1/accounting/settlement/statements/:id (access: workspace:books)

delete_accounting_trust_entries_id_match

ChatGPT
Unmatch a trust entry Removes an entry's bank-line match; the entry itself stays. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. DELETE /v1/accounting/trust/entries/:id/match (access: workspace:books)

delete_boards_id

ChatGPT
Delete a board and everything on it Destroys a custom board: its columns, options, rows, saved views, action buttons, grants and API keys all CASCADE away with it, and afterwards nothing can say what was there. No archive, no undo. To clear the records but keep the board, use post_boards_id_rows_delete. Manage on the board. DELETE /v1/boards/:id (access: board:manage)

delete_boards_id_actions_actionId

ChatGPT
Remove a board's action button Takes one button off a custom board's action bar for good. To withdraw a button reversibly, patch it with enabled: false instead. Manage on the board. DELETE /v1/boards/:id/actions/:actionId (access: board:manage)

delete_boards_id_column-policies_column

ChatGPT
Unlock a column on a board Clears one column's lock: everyone who can see the board sees that column again and may write it. Unlocking what was never locked succeeds — the caller asked for a state and that state now holds. Workspace admin plus manage on the board, Enterprise. DELETE /v1/boards/:id/column-policies/:column (access: workspace:admin)

delete_boards_id_columns_columnId

ChatGPT
Take a column off a board Removes one column from a custom board's shape. It is the SHAPE that goes: no row's stored cells are rewritten, and a column added back gets a new id, so the old values are unreachable rather than restored. Manage on the board. DELETE /v1/boards/:id/columns/:columnId (access: board:manage)

delete_boards_id_columns_columnId_options_optionId

ChatGPT
Remove a choice from a board column Takes one option out of a select column's vocabulary. Rows whose cell already holds that value keep it — this edits what may be CHOSEN, not what is stored. Manage on the board. DELETE /v1/boards/:id/columns/:columnId/options/:optionId (access: board:manage)

delete_boards_id_grants_accountId

ChatGPT
Stop sharing a board Takes one colleague's grant off a custom board. Nothing on the board changes and they keep whatever the workspace itself gives them; what goes is their reach into this board. Manage on the board. DELETE /v1/boards/:id/grants/:accountId (access: board:manage)

delete_boards_id_views_viewId

ChatGPT
Delete a saved board view Removes one saved view from a custom board; the rows it narrowed are untouched. The caller's own user view, or — for someone who manages the board — a workspace view. Another person's user view answers 404 rather than being deleted. DELETE /v1/boards/:id/views/:viewId (access: board:read)

delete_calendar_events_id

ChatGPT
Call a meeting off Calls off a meeting this service BOOKED (origin booked): it is removed from Google or Outlook and our row is marked cancelled; the row stays, so the record goes on saying a meeting was booked and then called off. A LINKED meeting (origin linked, the person's own event) is only taken OFF THE RECORD: our row is removed and their calendar is untouched, so do not say it was called off. Only the person whose calendar it is can do either; anything else is a 404. DELETE /v1/calendar/events/:id (access: workspace)

delete_comms_conversations_id_labels_labelId

ChatGPT
Take a label off a conversation Removes one inbox label from a thread the caller can read. The label itself stays. DELETE /v1/comms/conversations/:id/labels/:labelId (access: workspace)

delete_comms_conversations_id_participants_userId

ChatGPT
Take somebody off a conversation Removes one colleague from a thread's roster. A soft leave: who was on the thread when a message landed is part of what that message means, so nothing already sent changes. DELETE /v1/comms/conversations/:id/participants/:userId (access: workspace)

delete_comms_conversations_id_records_rowId

ChatGPT
Unfile a conversation from a record Stops a thread being filed under one record, so it leaves that record's Communications tab. The messages, the roster and the thread survive untouched — the address is what the conversation IS, and the record was only where it was filed. DELETE /v1/comms/conversations/:id/records/:rowId (access: workspace)

delete_comms_inboxes_medium_identifier_members_accountId

ChatGPT
Take somebody off an inbox Removes one account from the people who staff a number or a mailbox, which is what decides who may read its threads. Answers 204 whether or not they were on it. An admin for any inbox; the owner for their own connected mailbox or Zoom line. DELETE /v1/comms/inboxes/:medium/:identifier/members/:accountId (access: workspace)

delete_comms_labels_labelId

ChatGPT
Delete an inbox label Removes an inbox label. The threads it was on lose it and are otherwise untouched. DELETE /v1/comms/labels/:labelId (access: workspace)

delete_comms_messages_id

ChatGPT
Delete a message Retracts one message from a thread. A tombstone rather than a delete, so the transcript's shape stays honest. The author may remove their own; anybody else needs the thread's manage level and is refused with 403. DELETE /v1/comms/messages/:id (access: workspace)

delete_comms_numbers_id

ChatGPT
Release a phone number Give one bought line back to the carrier and mark it released, keeping the row so what was held between which dates stays answerable. Refused with 409 while it is the workspace's sending number. DELETE /v1/comms/numbers/:id (access: workspace:admin)

delete_comms_opt-outs_address

ChatGPT
Take an address off the do-not-contact list Removes one address from the do-not-contact list so sends to it are allowed again. Admins only. The address is in the path in any spelling; reason is required in the query and is written to the activity entry naming who removed it. An address not on the list is a 404. Only do this when the person has asked to hear from the workspace again. DELETE /v1/comms/opt-outs/:address (access: workspace:admin)

delete_comms_recordings_id

ChatGPT
Remove a call recording Delete one recording and its audio file. The call entry in the log stays, because the call still happened. DELETE /v1/comms/recordings/:id (access: workspace)

delete_comms_signatures_id

ChatGPT
Stop signing an address Removes the email signature from one sending address, by the signature row's id, so mail from that address goes out unsigned. Answers 204. DELETE /v1/comms/signatures/:id (access: workspace)

delete_credit-memos_id

ChatGPT
Delete a credit memo Deletes a credit memo, DRAFTS only: a submitted, approved or rejected memo reads as a not-found, so a decision trail cannot be erased. Answers 204. DELETE /v1/credit-memos/:id (access: workspace)

delete_definitions_key_actions_actionId

ChatGPT
Remove a row action from a page Removes a button from a record page's action bar for good. To take one out of use reversibly, patch it with enabled: false instead. Workspace admin. DELETE /v1/definitions/:key/actions/:actionId (access: def:admin)

delete_definitions_key_column-policies_column

ChatGPT
Clear a column's lock Unlocks one column: everyone who can see the page sees the column again and may write it. Unlocking what was never locked is a 204, not a 404 — the caller asked for a state and that state now holds. Workspace admin, and on the enterprise tier like the lock itself. DELETE /v1/definitions/:key/column-policies/:column (access: def:admin)

delete_definitions_key_columns_columnId

ChatGPT
Remove a column from a page Takes a column off a record page's shape. It is the SHAPE that goes: nothing rewrites the rows' stored cells, and a column added back gets a new id, so the old values are unreachable rather than restored. A column id that is not on this page is a 404. Workspace admin, and there is no archive for a column — to hide one from the default table instead, patch defaultEnabled to false. DELETE /v1/definitions/:key/columns/:columnId (access: def:admin)

delete_definitions_key_columns_columnId_options_optionId

ChatGPT
Remove a choice from a page's column Takes one option out of a select column's vocabulary AND clears it from every row holding it, so no cell keeps a value the column no longer names. Workspace admin. DELETE /v1/definitions/:key/columns/:columnId/options/:optionId (access: def:admin)

delete_definitions_key_credit-records_family_id

ChatGPT
Delete a credit record Deletes one credit record from a deal's underwriting tab. :family is credit_memo, decision or business_research and anything else is a 404; :id names the record itself, so the deal is not repeated in the path. A credit MEMO goes only while it is still a draft, and only for a ledger writer — once it has been through review it is the record of a decision and stays — while a decision or a piece of business research is a real delete, with the timeline entry as what survives. Deals page only. DELETE /v1/definitions/:key/credit-records/:family/:id (access: def:read)

delete_definitions_key_entity-groups_groupId

ChatGPT
Dissolve a relationship Removes the entity group and keeps every record in it. Use it when a grouping turns out to be wrong; the companies are untouched and are free to join another. DELETE /v1/definitions/:key/entity-groups/:groupId (access: def:read)

delete_definitions_key_entity-groups_groupId_members_rowId

ChatGPT
Take one record out of a relationship Removes one member. If it was the principal, the group is left with none rather than with a principal that is no longer a member. DELETE /v1/definitions/:key/entity-groups/:groupId/members/:rowId (access: def:read)

delete_definitions_key_grants_accountId

ChatGPT
Give a member a pack page back Removes a member's pack-page level, returning them to the default: every seat reads and writes the page. Admin seat, Enterprise. DELETE /v1/definitions/:key/grants/:accountId (access: def:admin)

delete_definitions_key_groups_groupId

ChatGPT
Delete a record group Dissolves a record group and its membership. THE RECORDS STAY on the board — the delete stops at the membership rows and has deliberately no path onward to the records themselves — and the timeline entry says how many were released. Use it when a grouping turns out to be wrong. DELETE /v1/definitions/:key/groups/:groupId (access: def:read)

delete_definitions_key_rows_rowId_affiliate-invitations_invitationId

ChatGPT
Revoke an outside affiliate's link Stamps one affiliate's link dead. Revoking twice is not an error. The caller must see the record and be able to edit its board. DELETE /v1/definitions/:key/rows/:rowId/affiliate-invitations/:invitationId (access: def:read)

delete_definitions_key_rows_rowId_api-connection

ChatGPT
Disconnect a lender's API Takes the API connection and its stored credentials off a Lenders record. Submissions already sent keep the id the lender's API gave them; the next send to a lender whose record names a portal or API is refused lender_api_not_connected until it is connected again. 204. Needs the seat that may edit the lender. DELETE /v1/definitions/:key/rows/:rowId/api-connection (access: def:read)

delete_definitions_key_rows_rowId_assignments_role_accountId

ChatGPT
Take one person off a role on a record Takes ONE person out of ONE role on a record, leaving the rest of that role's set standing, and answers the diff. Its own door rather than a PUT of the remaining list, because two people each removing a different colleague would otherwise overwrite each other. If the person removed was the one the record's cell named, the cell moves to whoever remains rather than being cleared. A person who does not hold the role is a 404. Many at once: post_definitions_key_rows_assignments DELETE /v1/definitions/:key/rows/:rowId/assignments/:role/:accountId (access: def:read)

delete_definitions_key_rows_rowId_books_entries_entryId

ChatGPT
Remove a client's journal entry Takes one entry and its lines out of the record's journal. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. DELETE /v1/definitions/:key/rows/:rowId/books/entries/:entryId (access: def:read)

delete_definitions_key_rows_rowId_books_rules_ruleId

ChatGPT
Forget a client's categorization rule Removes one rule. The lines it filed go to the next rule that matches, or back to review. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. DELETE /v1/definitions/:key/rows/:rowId/books/rules/:ruleId (access: def:read)

delete_definitions_key_rows_rowId_condition-shares_shareId

ChatGPT
Revoke a lender checklist link Takes one checklist link down now: it is stamped revoked in place and stops opening at once, and the row stays so the record of what was shared survives. Revoking twice succeeds. Any seat that writes the deal may revoke, not only whoever minted it. A link of another deal is a 404. DELETE /v1/definitions/:key/rows/:rowId/condition-shares/:shareId (access: def:read)

delete_definitions_key_rows_rowId_feed_reactions

ChatGPT
Remove a reaction from a record's feed Take the caller's own reaction off one feed item, named by the itemKind and itemId query parameters. Clearing a reaction that is not there succeeds. DELETE /v1/definitions/:key/rows/:rowId/feed/reactions (access: def:read)

delete_definitions_key_rows_rowId_folders_folderId

ChatGPT
Delete an empty document folder Remove a folder from a record. Refused with a conflict while the folder still holds a file or a sub-folder — move or delete what is inside it first. DELETE /v1/definitions/:key/rows/:rowId/folders/:folderId (access: def:read)

delete_definitions_key_rows_rowId_follow

ChatGPT
Unfollow a record End the caller's subscription to a record's feed. Not following is already the state asked for, so this succeeds either way. DELETE /v1/definitions/:key/rows/:rowId/follow (access: def:read)

delete_definitions_key_rows_rowId_merchant-invitations_invitationId

ChatGPT
Withdraw a merchant invitation Takes a merchant's link down: the invitation is stamped dead in place and stops working, and the row STAYS so the record of what was shared survives. Revoking twice succeeds. Any seat that can open the deal may revoke, not only the person who minted it. An invitation belonging to another deal is a 404. DELETE /v1/definitions/:key/rows/:rowId/merchant-invitations/:invitationId (access: def:read)

delete_definitions_key_rows_rowId_plaid_items_itemId

ChatGPT
Unlink a bank from a deal Invalidates the access token at Plaid first, then deletes the connection, its accounts, its transactions and every statement month computed from them. Use it when a merchant withdraws consent or a bank was linked to the wrong deal; the processor-read statements on the deal are untouched. DELETE /v1/definitions/:key/rows/:rowId/plaid/items/:itemId (access: def:read)

delete_definitions_key_rows_rowId_script

ChatGPT
Remove a record's call script Clears ONE lead's own script, so the page's status script or the workspace default answers for it again. Nothing is deleted — the version is retired, because what a rep was told to say on a call that went badly is evidence. A lead with no script of its own is a 404 rather than a silent success. Leads page only. DELETE /v1/definitions/:key/rows/:rowId/script (access: def:read)

delete_definitions_key_rows_rowId_subrecords_linkId

ChatGPT
Detach a subrecord Remove the link between a parent record and one of its children. THE CHILD RECORD IS NOT DELETED — it stays exactly where it was on its own page. Use the child page's own delete route to delete the record itself. DELETE /v1/definitions/:key/rows/:rowId/subrecords/:linkId (access: def:read)

delete_definitions_key_rows_rowId_super-kyc_id

ChatGPT
Stop a Super KYC link Stops watching: revokes the merchant's grants at Plaid and Google first (a Microsoft mailbox's token is discarded — Microsoft has no per-app revoke), deletes the bank rows, then marks the link stopped. The link and its findings stay as the record of what was watched. A provider that cannot be reached answers 502 and nothing is stopped. DELETE /v1/definitions/:key/rows/:rowId/super-kyc/:id (access: def:read)

delete_definitions_key_rows_rowId_vendor-slots_kind

ChatGPT
Empty one vendor slot Leaves the record with no vendor of that kind. Nothing is retired and the vendor itself is untouched. DELETE /v1/definitions/:key/rows/:rowId/vendor-slots/:kind (access: def:read)

delete_definitions_key_scripts_scriptId

ChatGPT
Remove a call script from a page RETIRES a call script rather than deleting it: nothing in this table is ever deleted, because what a workspace was telling its reps to say is evidence. Retiring the last live script for a target simply lets the next line of the resolution order — status, then workspace, then the built-in script — answer instead. Answers {retired: true}. Leads page, admin. DELETE /v1/definitions/:key/scripts/:scriptId (access: def:admin)

delete_definitions_key_statuses_statusId

ChatGPT
Remove a status from a page Removes a status from a page's vocabulary outright. Prefer patch_definitions_key_statuses_statusId with isActive: false when the workspace has simply stopped using the word: that keeps the records already standing on it and can be undone, while this leaves nothing behind. A status id not on this page is a 404. Workspace admin. DELETE /v1/definitions/:key/statuses/:statusId (access: def:admin)

delete_definitions_key_threads_threadId

ChatGPT
Delete a thread Deletes an internal thread and every message and participant on it, permanently. Reach for post_definitions_key_threads_threadId_resolve when a conversation is simply finished: resolving keeps the transcript readable and can be reopened, and this cannot. Answers 204. DELETE /v1/definitions/:key/threads/:threadId (access: def:read)

delete_definitions_key_threads_threadId_messages_messageId

ChatGPT
Delete a thread message Retracts one message the CALLER wrote; anyone else's is a 404. Soft: the row keeps its stamps so the transcript's shape and the pinned strip stay honest, and the words and the mentions are cleared. Answers 204. DELETE /v1/definitions/:key/threads/:threadId/messages/:messageId (access: def:read)

delete_definitions_key_threads_threadId_participants_userId

ChatGPT
Take someone off a thread Takes one colleague off a thread. Their messages stay in the transcript with their name on them, so nothing is lost — only the seat goes, and with it the notifications. Somebody who is not on the thread is a 404. Answers 204. DELETE /v1/definitions/:key/threads/:threadId/participants/:userId (access: def:read)

delete_definitions_key_views_viewId

ChatGPT
Delete a saved view Deletes a saved view: the caller's own user view, or — for a workspace admin — a workspace view. Another person's user view answers 404 rather than being deleted. The rows the view narrowed are untouched. Answers 204. DELETE /v1/definitions/:key/views/:viewId (access: def:read)

delete_files_id

ChatGPT
Delete a file Soft-delete one file by id: it stops being listed and stops being downloadable, while the row stays so its storage key can never be re-minted. The deletion is filed on the workspace's activity feed with the file's name and size. Works for any file in the workspace, loose or pinned to a record. DELETE /v1/files/:id (access: workspace)

delete_files_id_portal

ChatGPT
Hide a file from the merchant portal Stop showing one file of a deal to its merchant, answering the file with sharedToPortal false. The merchant's own uploads and the files that answered an ask still show; this takes back only what the desk shared. Hiding what is not shown is the same answer. DELETE /v1/files/:id/portal (access: workspace)

delete_files_id_shares_shareId

ChatGPT
Revoke a file's share link End one share link of a file now, by the id the share answered. Anyone who can open the file may revoke its links, not only whoever made them. The link stops working at once; revoking a link that already ended (revoked, or its file deleted or replaced) is the same answer, {revoked: true}. Filed on the activity feed. DELETE /v1/files/:id/shares/:shareId (access: workspace)

delete_files_id_star

ChatGPT
Unstar a file Clear the caller's own star on one file. Answers {starred: false}, also when there was no star. DELETE /v1/files/:id/star (access: workspace)

delete_forms_id

ChatGPT
Delete a form and its answers Destroys the form AND every submission filed through it, which is why the door is an admin's. When a form is simply finished, post_forms_id_unpublish stops the link and keeps the answers. DELETE /v1/forms/:id (access: workspace:admin)

delete_integrations_ringcentral_connection

ChatGPT
Disconnect RingCentral Disconnect the caller's own RingCentral extension and stop its history sync. Previously imported conversations and inbox ownership remain. Does not disconnect colleagues. Answers 204. DELETE /v1/integrations/ringcentral/connection (access: workspace)

delete_lender-workspaces_rowId

ChatGPT
Unlink a lender record from its Specter workspace Takes the Specter workspace off one entry on the Lenders board, so submissions to that lender stop going out in-network. rowId is the Lenders record. The RECORD stays — a lender we no longer submit to through Specter is still a lender — and shares already made carry their receiving workspace on the share itself, so nothing that happened is unmade here. put_lender-workspaces_rowId is how the link is made or corrected. DELETE /v1/lender-workspaces/:rowId (access: workspace)

delete_message-automation-rules_id

ChatGPT
Delete a standing message rule Removes one standing rule for good. To stop it firing without losing what it said, patch it with enabled: false instead. DELETE /v1/message-automation-rules/:id (access: workspace)

delete_message-drafts_id

ChatGPT
Delete a draft Removes one draft from the queue for good — the tidy-up, for a row nobody will ever act on. post_message-drafts_id_cancel is the verb that keeps the record of the decision and its reason; this one leaves nothing behind. DELETE /v1/message-drafts/:id (access: workspace)

delete_message-templates_id

ChatGPT
Delete a message template Removes one template for good. Refused while a draft or an automation rule still points at it — the answer names what is still using it. To retire a template without breaking those, patch it with isActive: false instead, which takes it out of the picker and keeps it readable. DELETE /v1/message-templates/:id (access: workspace)

delete_neo_memories_id

ChatGPT
Forget something Neo remembers Forgets one of the caller's memories by its id (read the list first) — "forget that", "stop doing that", "you don't need to remember that any more". Answers 204, and no later conversation starts with it. Another person's id is a 404. DELETE /v1/neo/memories/:id (access: workspace)

delete_notification-types_key

ChatGPT
Retire a notification type Removes one of the workspace's OWN notification types by its key, taking its row off everybody's notification settings. One of Specter's own types is refused with a sentence saying to switch it off instead. Admin only. DELETE /v1/notification-types/:key (access: workspace:admin)

delete_packs_key_samples

ChatGPT
Remove a pack's sample record Removes every row the pack's sample made that is still there, the last made first, through the same delete a person's row delete goes through — nothing else. A row this seat may not delete refuses the whole removal. Answers the pack and how many rows went (0 when the sample is already gone). DELETE /v1/packs/:key/samples (access: workspace)

delete_payment-schedules_id

ChatGPT
Destroy a schedule and its instalments Really deletes one schedule and every instalment and allocation under it, in one transaction. Answers 204. Reach for it only for a schedule that should never have existed: to stop collecting on a live one use post_payment_schedules_id_pause, and to close out one that ran its course use post_payment_schedules_id_complete. Recorded cash flows SURVIVE with their link cut, because money that moved stays a fact, and the deletion is filed on the record's timeline with what it destroyed. A schedule belonging to another workspace deletes nothing and answers not found. DELETE /v1/payment-schedules/:id (access: workspace)

delete_reports_shares_id

ChatGPT
Revoke a public report link Revokes one published public link: the URL stops working immediately for everybody holding it, and a stranger who visits it gets a flat not-found. The share row itself is kept, so "this report was on a public screen until March" stays answerable. Any admin, not only whoever published it. This leaves the report alone; delete_reports_views_id deletes the report and all of its links. DELETE /v1/reports/shares/:id (access: workspace:admin)

delete_reports_views_id

ChatGPT
Delete a saved report DESTRUCTIVE: deletes a saved report, and its published public links go with it — anybody holding one of those URLs loses the page. The owner's or an admin's; a built-in cannot be deleted. To take down one published link and keep the report, delete_reports_shares_id. DELETE /v1/reports/views/:id (access: workspace)

delete_table-preferences_tableKey

ChatGPT
Remove table-preferences tableKey Drops the CALLER's own saved layout for one grid — tableKey namespaced like table:deals:overview — so that grid falls back to the workspace default. Deleting what was never saved is still a no-content: the caller asked for a state and that state now holds. Written by the grid's own Reset; get_table-preferences_tableKey reads the layout back. DELETE /v1/table-preferences/:tableKey (access: workspace)

delete_webhooks_id

ChatGPT
Remove webhooks id Remove a subscription; nothing is delivered to it again and its secret is gone. Reach for PATCH with status 'paused' instead when the workspace wants deliveries to stop but the row, its history and its secret to survive. Admin only. DELETE /v1/webhooks/:id (access: workspace:admin)

delete_workflows_id

ChatGPT
Delete an unpublished draft DESTRUCTIVE and permanent: deletes a workflow draft that never published. A workflow with any version is refused — its runs are the record of a machine acting inside the workspace and they are not allowed to go. Retire a published rule with patch_workflows_id {"status": "archived"} instead. DELETE /v1/workflows/:id (access: workspace:admin)

delete_workspace_compliance_reports_id

ChatGPT
Delete a compliance report Admin only, and the one destructive act in this family: every run of the report goes with it, sign-offs included. Retract a question only when it should never have been asked; to stop asking it while keeping the evidence, PATCH the report with active set to false. DELETE /v1/workspace/compliance/reports/:id (access: workspace:admin)

delete_workspace_domains_id

ChatGPT
Remove a workspace domain Gives up a claimed domain. Refused with 409 while any inbound address still receives at it - delete those addresses first. DELETE /v1/workspace/domains/:id (access: workspace:admin)

delete_workspace_domains_id_form-host

ChatGPT
Remove the default custom form address Removes the workspace's own form address under this domain, so form links are composed on the default address again. Links already sent on the old address keep opening for as long as its CNAME stands. 404 when no form address is set under the domain. DELETE /v1/workspace/domains/:id/form-host (access: workspace:admin)

delete_workspace_drive-watches_watchId

ChatGPT
Stop watching a Drive folder Stop watching a folder. There is no pause and no disable - this is the whole other end of the lifecycle. Documents the watch already landed stay where they are, and so does the record of which Drive files came out of that folder, so re-adding the same folder later does not land everything in it a second time. DELETE /v1/workspace/drive-watches/:watchId (access: workspace:admin)

delete_workspace_inbound_id

ChatGPT
Remove an inbound address Stops an address receiving and forgets it. Mail already filed through it keeps its records: a message's record is a stored fact, not a re-run of this rule. DELETE /v1/workspace/inbound/:id (access: workspace:admin)

delete_workspace_invites_id

ChatGPT
Withdraw an invite Revoke an invite so its link stops working. Admin only, and a not-found when the id is not this workspace's. It does not touch a seat somebody already accepted into — that is DELETE /v1/workspace/members/:accountId. DELETE /v1/workspace/invites/:id (access: workspace:admin)

delete_workspace_members_accountId

ChatGPT
Remove a member from the workspace Take a seat away and settle what pointed at it: board grants are deleted, team rows go (teams they LED survive, leaderless), their tasks and rows become unassigned, and boards they made stand. ACCESS GOES, WORK STAYS — nothing is deleted because somebody left, and the answer counts each. Admin only; an owner can only be removed by an owner, and never the last one (409). DELETE /v1/workspace/members/:accountId (access: workspace:admin)

delete_workspace_processor_aliases_aliasId

ChatGPT
Remove an alias from the workspace's intake mailbox Stops treating an address as one of the workspace's processor accounts. The address stays at Google. DELETE /v1/workspace/processor/aliases/:aliasId (access: workspace:admin)

delete_workspace_processor_file-types_key_examples_exampleId

ChatGPT
Remove an example from a file type Takes one example document off a file type in processor-v2's catalogue. :key is the app's document word (bank_statement, id, …) or a workspace's own custom_<slug> type; :exampleId is the example's id (its sha256) from the list. 400 for a word processor-v2 has no type for; 501 while processor-v2 does not serve removing an example; 503 when this deployment has no processor-v2. Admin only. DELETE /v1/workspace/processor/file-types/:key/examples/:exampleId (access: workspace:admin)

delete_workspace_processor_whitelist_entryId

ChatGPT
Remove a sender from the workspace's processor whitelist Takes one of this workspace's whitelist entries off, with ?reason= saying why. The entry is kept with who removed it and why. 404 when no live entry of this workspace has that id. Admin only. DELETE /v1/workspace/processor/whitelist/:entryId (access: workspace:admin)

delete_workspace_role-presets_presetId

ChatGPT
Remove a role preset Removes a preset. Nobody loses access: applying it wrote each member's own seat and grants. 409 when an outside affiliate was invited under it — that preset is kept with the invitation for the audit. A workspace admin's. DELETE /v1/workspace/role-presets/:presetId (access: workspace:admin)

delete_workspace_sidebar-folders_id

ChatGPT
Delete a sidebar folder Deletes a folder for the whole workspace. What was filed in it returns to the rail in its own place; nothing it pointed at is touched. Recorded on the workspace's activity feed. Any seat. DELETE /v1/workspace/sidebar-folders/:id (access: workspace)

delete_workspace_sla-policies_id

ChatGPT
Withdraw a stage's SLA Admin only. The breaches recorded against the policy go with it, because a breach is only meaningful against the target it broke. To stop measuring while keeping the history, PATCH the policy with active set to false instead. DELETE /v1/workspace/sla-policies/:id (access: workspace:admin)

delete_workspace_teams_teamId

ChatGPT
Delete a team Dissolve a team. The people keep their seats and their work — a team is a grouping and a reporting line, not access. Admin only. DELETE /v1/workspace/teams/:teamId (access: workspace:admin)

delete_workspace_teams_teamId_members_accountId

ChatGPT
Take a member off a team Take one member off one team. Their seat, their access and their work are untouched; this only changes whose work the team's lead is answerable for. Admin only. DELETE /v1/workspace/teams/:teamId/members/:accountId (access: workspace:admin)

delete_workspace_underwriting_rule-sets_setId_rules_ruleId

ChatGPT
Delete an underwriting rule Remove one draft rule. Past evaluations keep their own copy of every rule's words, so nothing they explain is lost. DELETE /v1/workspace/underwriting/rule-sets/:setId/rules/:ruleId (access: workspace:admin)

delete_workspace_vendors_id

ChatGPT
Remove a vendor Admin only. Every deal slot that named this vendor goes with it. When the deals that used it are still open, retire it instead by PATCHing it with active set to false. DELETE /v1/workspace/vendors/:id (access: workspace:admin)

delete_workspace_vendors_id_documents_fileId

ChatGPT
Unpin a vendor's document Admin only. Takes the reference off; the FILE stays in the workspace's library. Unpinning is "this is not the engagement letter", never "delete the document". DELETE /v1/workspace/vendors/:id/documents/:fileId (access: workspace:admin)

execute_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Applies an already-accepted Heartbeat action — performs the field update or sends the prepared message. Executing an already-applied action is an idempotent no-op (already-sent work is never repeated). Confirm with the user first; this writes live data. Field updates can be undone with revert_heartbeat_action, but outbound sends cannot. Operates on one stored, accepted proposal by ID; it does not accept arbitrary operation names or register additional model-callable tools.

file_support_ticket

ChatGPT
Files a platform support ticket as the authenticated user — the same reporter-side ticket system as the in-app "report a problem" flow. USE THIS whenever the platform itself misbehaves during a conversation: an endpoint returns wrong or oversized data, a filter is ignored, a tool result truncates so a task cannot be completed, or the user reports something broken. Include what was attempted, the exact call that misbehaved, and what came back wrong in the body. The ticket lands with the Specter system admins; identity (tenant, reporter) is stamped server-side. Tell the user the ticket reference from the response. This creates a record but sends nothing to any customer.

find_my_free_slots

ChatGPT
For non-merchant Specter broker/operator users only. The signed-in user's OPEN windows — their working hours in their own time zone minus what their calendar (Google or Outlook) already holds — for a day or a span of days. Answers "when am I free tomorrow?", "find me 45 minutes with the merchant this week", and is what to call BEFORE proposing a meeting time (propose_meeting) so the time is one the person can take. Give day (today / tomorrow / a weekday name / YYYY-MM-DD) and optionally days=N (up to 14), duration (minutes; default: their default meeting length) and limit. Each slot is a start, an end and its length; the answer also carries the working day it was cut from. When connected is false, repeat the reason to the user verbatim. Read-only.

get_accounting_access

ChatGPT
List accounting grants Every seat holding an accounting grant — the people besides owners and admins who may open the brokerage's books: who, their seat role, the level (read or manage), and whether the grant reaches every linked bank account or only the ones listed. The admin seat. GET /v1/accounting/access (access: workspace:admin)

get_accounting_allocations

ChatGPT
List allocations What was allocated against what: each row says this much of that bank line was this deal's commission, PSF, clawback recovery, bonus, management fee or syndication fee. These rows are the SOLE source of every "collected" number in accounting, so this is the read for "what did we allocate against that deal". Filters: ?dealId= (the deal's own reconciliation), ?bankTransactionId= (what one bank line paid for), ?limit=. The answer also carries types, the closed vocabulary a new allocation may use. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. GET /v1/accounting/allocations (access: workspace:books)

get_accounting_bank-transactions

ChatGPT
List bank transactions The bank lines imported into the workspace — money that has actually moved — each with its source, posted date, amount, account name and last four, description and memo, beside how much of it has been allocated to a deal and how much is still unclaimed. Filters: ?source=, ?direction=, ?unreconciled=true (only lines with money on them nobody has claimed, which is the reconcile-the-bank worklist), ?from= / ?to= on the posted date, ?limit=. What a line actually paid FOR is the allocations: get_accounting_allocations with ?bankTransactionId=. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. GET /v1/accounting/bank-transactions (access: workspace:books)

get_accounting_bank-transactions_id

ChatGPT
Read a bank transaction One imported bank line — source, posted date, amount, account name and last four, description and memo — with how much of it is allocated, how much is unallocated and whether it is fully allocated. A cost, an allocation and an exception have no by-id read of their own — those are read through their lists. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. GET /v1/accounting/bank-transactions/:id (access: workspace:books)

get_accounting_categories

ChatGPT
Read workspace transaction categories Read the industry-neutral category catalogue, including stable identifiers, types, matching keywords, column tags and guidance for Neo. Read this before categorizing transactions. Archived categories remain available as historical labels. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. GET /v1/accounting/categories (access: workspace:books)

get_accounting_company-accounts

ChatGPT
List company bank accounts and their purposes Every company bank account with its purpose — operating, trust (IOLTA) or escrow — the packs (practice areas) a trust account serves (empty = firm-wide), and its bank balance (Plaid's for a linked account, the typed statement for a manual one). To an admin it also lists linked company accounts no purpose has been set for (unregistered), and the packs installed here to tag with. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/company-accounts (access: workspace:books)

get_accounting_company-accounts_id_reconciliation

ChatGPT
Three-way reconciliation of a trust account For one trust or escrow account: the bank balance, deposits in transit and outstanding disbursements (journal entries not yet matched to a bank line), the adjusted bank balance, the trust journal's total, the sum of the client-matter ledgers, whether all three agree (balanced, absent with no bank figure), each matter's ledger, and the unmatched entries. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/company-accounts/:id/reconciliation (access: workspace:books)

get_accounting_dashboard

ChatGPT
Read the accounting dashboard The workspace's money at a glance: how many funded deals, what was funded, the commission and PSF agreed, how much has actually been paid (collected), what is still owed (outstanding), costs, open clawback, what every participant could draw right now (available), what is held, pending payouts and their total, deals on hold, defaulted or past their SLA, and how many bank lines are still unreconciled. ?from= and ?to= (dates) narrow the deal counts to a window; absent means all of it. This is the summary strip — for the deal-by-deal list reach for get_accounting_deals, and for what is WRONG rather than what is totalled, get_accounting_exceptions. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. This is the commission ledger — only deals booked into accounting, with their commissions, payouts and what was collected. How many deals funded and for how much, by lender, stage or period, is the deals page's ONE read: get_definitions_key_aggregates on deals with groupBy=Lender and calculate=Funded:sum, narrowed by the Funded Date column. GET /v1/accounting/dashboard (access: workspace:books)

get_accounting_deal-costs

ChatGPT
List deal costs What funded deals cost the house: each cost's category, description, vendor, amount and status (open, paid or void). Costs reduce net revenue, so they move every net-basis participant's availability. Filters: ?dealId=, ?status=, ?limit=. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. GET /v1/accounting/deal-costs (access: workspace:books)

get_accounting_deal-splits

ChatGPT
List deal splits Who is owed a share of a funded deal, and how much: per stake, the party and its kind, the basis and the share, what they are entitled to, how much of it has actually been paid in, what they have drawn, what is pending, their share of any open clawback, whether they are held, and what is available to draw right now. Filters: ?dealId=, ?accountId=, ?crmUserId=, ?partyKind=, ?status=, ?limit=. This is the whole book, so it is how to answer "what is this rep owed" (?crmUserId=) and "who are the syndicators on this deal" (?partyKind=syndicator) — there is no separate syndications route. For the CALLER's own drawable balance, get_accounting_me_availability. WHOSE stakes come back is decided by the caller, not the filter: the books read every participant; a member reads only their own stake and a team lead their own and their reps'. A filter naming somebody outside that reach returns nothing. GET /v1/accounting/deal-splits (access: workspace)

get_accounting_deal-splits_id

ChatGPT
Read a deal split One participant's stake on one funded deal: the party, its kind, the basis, the share or flat fee, its status and hold, with entitled, collected, drawn, pending, clawback share and what is available to draw. Somebody else's stake is a not-found unless the caller holds the books or leads its holder. GET /v1/accounting/deal-splits/:id (access: workspace)

get_accounting_deals

ChatGPT
List deals in the ledger One page of the commission ledger: per funded deal, the agreed economics (funded amount, commission, PSF) beside the rollup — what has actually been paid against it, what is still owed, costs, open clawback, how long it has been aging and whether it is past its SLA. Filters: ?facilityId=, ?lenderId=, ?status=, ?reference=, ?onHold=, ?pastSla=, ?from= / ?to= on the deal date, ?limit=. Narrowing by a rep or a team lead is not here — a participant is a stake, so ask get_accounting_deal-splits with ?crmUserId=. For ONE deal with its stakes, costs, allocations and payouts in a single read, get_accounting_deals_id_detail. Answered in the CALLER's view, named in view: books (an admin or an accounting grant) is the whole agreement; own and team are a participant's — only deals they hold a stake on or are assigned to, with the house's side (commissionMinor, psfMinor, the rollup, costs, notes) left at zero and mine carrying their own stake's totals (a team lead's includes the reps they lead). A deal outside that reach is a not-found. This is the commission ledger — only deals booked into accounting, with their commissions, payouts and what was collected. How many deals funded and for how much, by lender, stage or period, is the deals page's ONE read: get_definitions_key_aggregates on deals with groupBy=Lender and calculate=Funded:sum, narrowed by the Funded Date column. Use ?page=1&limit=200 for a bounded numbered page with items, nextCursor, and the exact filtered total. page cannot be combined with cursor. ?paidOut=false excludes paid-out rows before paging; ?search= is a case-insensitive literal reference substring; repeated ?statuses= selects multiple status words. Page rows and count share a read snapshot. Alternatively ?paged=true&cursor= retains keyset traversal without a total. Bare calls retain the legacy bounded array. GET /v1/accounting/deals (access: workspace)

get_accounting_deals_id

ChatGPT
Read a ledger deal One funded deal's agreed economics with its rollup: funded amount, commission, PSF, hold state, default and clawback, SLA, and how much has actually been paid against it versus what is still owed. For the same deal WITH its stakes, costs, allocations and payout requests, reach for get_accounting_deals_id_detail instead — one round trip rather than five. Answered in the CALLER's view, named in view: books (an admin or an accounting grant) is the whole agreement; own and team are a participant's — only deals they hold a stake on or are assigned to, with the house's side (commissionMinor, psfMinor, the rollup, costs, notes) left at zero and mine carrying their own stake's totals (a team lead's includes the reps they lead). A deal outside that reach is a not-found. GET /v1/accounting/deals/:id (access: workspace)

get_accounting_deals_id_activity

ChatGPT
Read accounting correction history Read a booked deal’s tenant-scoped activity and correction reasons with before/after money values. Requires books read access. Follow nextBeforeId with beforeId to reach older history. GET /v1/accounting/deals/:id/activity (access: workspace:books)

get_accounting_deals_id_detail

ChatGPT
Read a ledger deal in full Everything hanging off one funded deal in a single read: the deal and its rollup, every stake (who is entitled to a share, and what each has collected, drawn and may still draw), its costs, the bank allocations that paid it, and its payout requests. What the deal's accounting panel opens on, and the closest thing here to "show me this deal's reconciliation". Not a snapshot: the five reads run in five transactions, so an allocation landing mid-request can show a rollup one row ahead of the list beneath it — stale, never wrong, so refetch rather than building an invariant on the five agreeing. Answered in the CALLER's view, named in view: books (an admin or an accounting grant) is the whole agreement; own and team are a participant's — only deals they hold a stake on or are assigned to, with the house's side (commissionMinor, psfMinor, the rollup, costs, notes) left at zero and mine carrying their own stake's totals (a team lead's includes the reps they lead). A deal outside that reach is a not-found. Without the books, splits holds only the caller's own stake (and a team lead's reps'), payouts only the caller's own requests, and costs and allocations are empty. GET /v1/accounting/deals/:id/detail (access: workspace)

get_accounting_exceptions

ChatGPT
List accounting exceptions What is wrong across the workspace's books, as five predicates over named columns rather than a heuristic: past_sla (revenue still owed past the deal's own SLA days), on_hold (a deal frozen, so nobody on it is being paid), no_splits (commission earned with nobody entitled to it), clawback_open (a raised clawback with nothing recovered against it) and unreconciled (a bank line with money on it nobody has claimed). Each item carries its kind, severity, subject, amount and how long it has stood; an exception has no id of its own because it is evaluated per read. ?limit= caps the list. Ask this for "what needs attention"; get_accounting_dashboard for the totals. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. GET /v1/accounting/exceptions (access: workspace:books)

get_accounting_invoice-lines_recent

ChatGPT
List recently billed invoice lines What this workspace has billed in the last year, each description once with the price and tax it last carried, most used first — the invoice editor's suggestions. Takes the brokerage's books; anyone else is refused with 403. GET /v1/accounting/invoice-lines/recent (access: workspace:books)

get_accounting_invoice-payments

ChatGPT
Read the workspace's online invoice payments Whether this workspace takes its clients' invoice payments online: available (this deployment can connect Stripe at all), connected, the connected Stripe account's id and name, whether it can take a charge (chargesEnabled) and has finished Stripe's onboarding (detailsSubmitted), and applicationFeeBps (Specter's share, 0 by default — the firm pays Stripe's fees). Re-reads the account's standing from Stripe. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant; anyone else is refused with 403. GET /v1/accounting/invoice-payments (access: workspace:books)

get_accounting_invoice-settings

ChatGPT
Read the workspace's invoicing setup The invoicing defaults from Accounting › Settings › Invoicing: the business name, logo and color on an invoice (as saved, and effective with the workspace profile's fallbacks), the number prefix and next number, the default payment terms in days (0 is due on receipt) and tax (taxPpm, 82500 is 8.25%), the default how-to-pay text, whether clients can pay online, a checklist of what is done and what is left, ready (the workspace can invoice), and a preview of the client's email and text filled with a sample invoice. Takes the brokerage's books; anyone else is refused with 403. GET /v1/accounting/invoice-settings (access: workspace:books)

get_accounting_invoices

ChatGPT
List client invoices Every invoice this workspace has drafted or issued to its customers, on any page's record, newest first: number, who it bills, total, paid, balance and status (draft | open | partially_paid | paid | overdue | void, derived from the invoice's journal), with what is outstanding, what is overdue, what was paid and aging buckets of what is still owed. Use it for "who owes us money", "what's overdue", "unpaid invoices", "accounts receivable", or to find one invoice by its number or customer before acting on it. Filters: ?status= (overdue for late ones), ?q= (an invoice number like INV-1003, the customer's name or email, or the record's name), ?aging= (current | 1-30 | 31-60 | over-60 days past due), ?unviewed=true (issued, owed and never opened by the client), ?rowId= (one record's). Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant; anyone else is refused with 403. GET /v1/accounting/invoices (access: workspace:books)

get_accounting_invoices_attention

ChatGPT
What the invoices need today The invoices' to-do list in one read: overdue (count, total owed and the invoices, each with days late and the last reminder's day), unviewed (issued before today and never opened by the client), staleDrafts (drafts more than a day old, never issued), paymentsSinceYesterday (payments booked since yesterday's start, reversals left out), outstandingMinor, and suggestions — one-press asks such as "Send reminders to the 3 overdue invoices", each with the words to act on. Use it first for "how are our invoices", "what needs following up", "did anyone pay", or a morning check-in on money owed; act through the remind, issue and payment tools. Optional ?timeZone= (IANA) cuts the day. Takes the brokerage's books; anyone else is refused with 403. GET /v1/accounting/invoices/attention (access: workspace:books)

get_accounting_invoices_id

ChatGPT
Read an invoice One invoice whole: its terms, lines, totals, journal (issued, payment, reversal, void entries — a payment's entry id is what reversing it names) and status, and — once issued — the client's link (link), where they view it, download the PDF and pay. Use it before editing a draft, recording or reversing a payment, or to answer what one invoice says. Takes the brokerage's books; anyone else is refused with 403. GET /v1/accounting/invoices/:id (access: workspace:books)

get_accounting_invoices_id_mail-preview

ChatGPT
Preview an invoice's email The email post.accounting.invoices.id.send would send for this invoice: to, from (the workspace's paper sender), subject and words, with an optional ?message= quoted in it. A draft previews with the number it will take and says its link is made when it is issued. 400 when the workspace has no address to send from. Takes the brokerage's books; anyone else is refused with 403. GET /v1/accounting/invoices/:id/mail-preview (access: workspace:books)

get_accounting_matters_rowId_settlement

ChatGPT
Read a matter's settlement Everything about one matter's settlement, the same whether opened from Accounting or from the case record: the current statement (draft, else approved) with lines, what is paid on each and the readiness checklist; its versions; the matter's trust ledger per account; its receivables; the trust accounts that may hold its money; the closing checks and closing statuses; and a suggestion read from the row (settlement amount, fee share, providers linked with their bills and liens). Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/matters/:rowId/settlement (access: workspace:books)

get_accounting_me_access

ChatGPT
Read the caller's accounting access What the CALLER may do with the brokerage's books right now: view (books, team or own), whether they read the whole book (readBooks), keep it (manageBooks), decide payouts and link banks (approve), and book deals and lay stakes (bookDeals); which linked bank accounts a grant reaches (allAccounts, bankAccounts); and the same standing as permission words. Read this first when somebody asks an accounting question: a seat with view: own is answered from get_accounting_deals, get_accounting_me_availability and get_accounting_payouts, and has no bank lines, costs or other people's commissions to be told about. GET /v1/accounting/me/access (access: workspace)

get_accounting_me_availability

ChatGPT
Read the caller's own available balance What the CALLER may draw right now, deal by deal: each of their own stakes with the deal's reference, the party kind and basis, entitled, drawn and available, whether it is held, and the total available across all of them. "What am I owed", in other words. It takes no subject on purpose — what one person is owed is not a question another member gets to ask, and an operator wanting the whole book asks get_accounting_deal-splits. A draw is against ONE stake, so the splitId here is what post_accounting_payouts needs. GET /v1/accounting/me/availability (access: workspace)

get_accounting_me_payout-methods

ChatGPT
List the caller's payout methods Where the CALLER's draws are sent: each method's kind (bank_account, zelle or other), label, bank, account type, last four and whether it is the default. Never the account number, routing number or handle — no read returns those to a list. It takes no subject: where somebody is paid is not a question one member gets to ask about another. GET /v1/accounting/me/payout-methods (access: workspace)

get_accounting_payouts

ChatGPT
List payout requests Draws against stakes and the decision on each: who asked, for which stake and deal, how much, what was available at the moment of the request, the status (pending, approved, rejected, paid or cancelled), who decided it and when, and a rejection's reason. Filters: ?splitId=, ?dealId=, ?status=, ?mine=true (the caller's own requests), ?limit=. What somebody COULD draw, as against what they have asked for, is get_accounting_me_availability. Without the books the answer is always the caller's own requests, whether ?mine= is sent or not. destination (kind, label, bank and last four — never the number) appears only to the requester and to an admin. GET /v1/accounting/payouts (access: workspace)

get_accounting_payouts_id

ChatGPT
Read a payout request One payout request: its stake and deal, the amount asked for, the balance that was available when it was made, its status, and the decision on it with the reason and who signed it. Somebody else's request is a not-found without the books. GET /v1/accounting/payouts/:id (access: workspace)

get_accounting_plaid-connections

ChatGPT
List company bank feeds Lists company banks and their account selection for this workspace administrator. Merchant and deal bank links are separate. Use to explain which accounts feed the accounting ledger. GET /v1/accounting/plaid-connections (access: workspace:admin)

get_accounting_profit-loss

ChatGPT
Generate cash profit and loss Generate a date-bounded, currency-specific cash P&L from categorized watched-account transactions. Group by day, week or month. Loan principal, transfers, equity, capital assets, uncategorized activity and provider changes awaiting review are excluded from profit and reported separately. Read-only; does not trigger a Plaid refresh. Use from and to dates, interval and currency. Review coverage and classification before using for financing. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. GET /v1/accounting/profit-loss (access: workspace:books)

get_accounting_receivables

ChatGPT
List receivables across every pack Money expected on any page's row — a carrier's settlement, a fee — across every pack: payer, expected amount and date, trust account, received so far and status (expected | received | disbursed | closed | cancelled, derived from the trust journal), with totals and ageing buckets of what is still expected. Filters: ?pack=, ?accountId=, ?status=, ?rowId=. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/receivables (access: workspace:books)

get_accounting_settings

ChatGPT
Read accounting settings Read this workspace's accounting settings: the default currency, the fiscal-year start month, how many days a payout may wait and an exception may sit before it is flagged, the SLA and clawback window a new deal gets when it does not say, and the wording of every closed vocabulary (allocation types, split party kinds, split bases, exception kinds with what each means and how it is fixed, bank sources, payout statuses). The answer is {settings, vocabularies}; vocabularies lists each vocabulary's fixed keys in display order; the keys never change, only their wording. GET /v1/accounting/settings (access: workspace)

get_accounting_settlement_statements_id_document

ChatGPT
Preview a closing statement The statement as a document: its named values (the settlement_statement template's placeholders), its line items, the template used, and a markdown preview with the figures filled. A draft's preview says DRAFT. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/settlement/statements/:id/document (access: workspace:books)

get_accounting_trust_entries

ChatGPT
List trust journal entries The trust journal, newest first: deposits and disbursements per client matter per trust account, with payee, method, check or wire reference, date, the statement line paid, any reversal and the bank line matched. Filters: ?accountId=, ?rowId=. The journal is append-only. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. An account-scoped grant reaches only the trust accounts on the linked bank accounts it names. GET /v1/accounting/trust/entries (access: workspace:books)

get_activity

ChatGPT
Read the workspace's activity feed The workspace activity log, paged, newest first: who did what to which record. Answers "what happened" and "who changed this". GET /v1/activity (access: workspace:admin)

get_agents_performance

ChatGPT
Agent performance Funded volume and conversion per member over a window, plus a funded-amount series by month. Two windows, named: the cohort is deals CREATED in the window with its funded share as the conversion rate, and the funded figures are deals FUNDED in the window whenever they were opened. A deal is attributed to its closer, falling back to its opener. Use it to answer who is closing and how much. GET /v1/agents/performance (access: workspace)

get_agents_rowId_credits

ChatGPT
Read what an agent brought in The leads and the deals filed against one agent on the agents registry — ISO agents and brokers' external reps. Two lists, because 'names handed over' and 'deals that survived' are the two numbers a commission conversation needs. GET /v1/agents/:rowId/credits (access: workspace)

get_application_details

ChatGPT
Gets one authenticated facility/application by ID, including currently available related entities and pipeline status. Use after list_my_recent_applications when the user asks about a specific application. Requires account authorization and does not mutate data.

get_application_history

ChatGPT
Reads timeline/history for a facility/application visible to the authenticated user, including related activity/events the backend exposes. Use when the user asks what happened on an application or wants recent changes. It is read-only.

get_assignments_for_entity

ChatGPT
For non-merchant Specter broker/operator users only. Reads who is currently assigned to one record, across every role, with each assignee's name and email resolved. Call this BEFORE set_assignments_for_entity: that tool REPLACES a role's whole assignee list, so adding a person means sending the existing ones back alongside them, and this is where you read them. Read-only.

get_bank_statement

ChatGPT
Gets one bank statement visible to the authenticated user. Use when the user asks for statement details, analysis, linked facility, risk signals, or included periods. It is read-only.

get_bank_statement_history

ChatGPT
Reads timeline/history for a bank statement visible to the authenticated user. Use when the user asks how a statement or its analysis changed. It is read-only.

get_bank_statement_period

ChatGPT
Gets one bank statement period visible to the authenticated user. Use for period-level revenue, balances, NSF, DSR, fraud, and MCA-position analysis. It is read-only.

get_billing

ChatGPT
Read the workspace's billing The Billing pane's whole subject in one read: the plan tier and its status, seats paid against seats in use, the current period, the next renewal and its amount, the card on file, and the AI credit balance. Reach for this to answer what a workspace is paying for; GET /v1/workspace/plan is the narrower read when only the entitlement matters. Carries no Stripe identifiers. GET /v1/billing (access: workspace)

get_billing_credits

ChatGPT
Read the AI credit balance and ledger The workspace's credit balance and the append-only ledger behind it, newest first: every purchase, grant, spend and adjustment with what it pointed at. The balance is the sum of the ledger, so the two can never disagree. Reach for this to answer where a workspace's credits went. GET /v1/billing/credits (access: workspace)

get_billing_invoices

ChatGPT
Read the billing history The workspace's recent Stripe invoices: date, amount in minor units, currency, status and a link to the hosted invoice page. A workspace that has never paid answers an empty list rather than a refusal. Needs the deployment's Stripe key; without one it answers 503 leading with billing_unconfigured. GET /v1/billing/invoices (access: workspace)

get_billing_usage

ChatGPT
Read this period's usage What the workspace consumed in the period it is paying for: seats in use against seats paid, credits spent and left, messages a person sent, and live file bytes held. The window is the subscription's own period, or the last thirty days when there is no subscription, and the answer states which. GET /v1/billing/usage (access: workspace)

get_blasts

ChatGPT
Read the blasts Every blast this workspace has — one set of words aimed at many people — text and email in one list, each with its name, kind, status (draft, scheduled, sending, sent, partial, cancelled), how many addresses are in its audience, what became of the copies, and the line they went out on. ?kind=sms|email, ?status=, ?limit=. get_blasts_stats counts the workspace's blasts by status for a header, and get_blasts_id_report opens one up recipient by recipient with their replies. GET /v1/blasts (access: workspace)

get_blasts_id_opens

ChatGPT
Read estimated campaign email opens Returns tracked sent email counts, messages with an observed image load, observation totals and first/last observation times for an accessible email blast. Use to assess Outreach engagement. Image loads are estimates affected by privacy proxies, caching and blocked images, not proof of reading. GET /v1/blasts/:id/opens (access: workspace)

get_blasts_id_preview

ChatGPT
Preview how a blast will land How a blast will land BEFORE it is sent, counted over its whole audience and rendered by the same function the send uses — read it after aiming a blast and before dispatching it. It answers wouldSend, how many resolve every field from its first choice (clean) or from a fallback (fellBack), how many would be skipped and why (unresolved: a field with nothing to fill it and no fallback; empty; optedOut), every field the words use with how many recipients can fill it (fields), up to limit (default 5, at most 25) real samples with the troubled ones first, and problems — sentences to fix before sending: a malformed field, a field nobody in the audience has, recipients who would be skipped, a missing goal. It sends nothing and changes nothing. Each sample's subject and body are that recipient's message as the dispatch writes it, fields filled — the copy to show a person reviewing the blast, never the {{…}} source. GET /v1/blasts/:id/preview (access: workspace)

get_blasts_id_report

ChatGPT
Read a blast's report One blast's whole picture in a single read: its counts (sent, replied, opted out, failed, skipped, queued over the frozen audience), the senders it went from, and a page of recipients with each one's outcome, sent time and the thread it opened. Use it for "what happened with that blast" and "who replied". GET /v1/blasts/:id/report (access: workspace)

get_blasts_merge-fields

ChatGPT
List the merge fields a blast can use The names a blast aimed at one board may be written with: ?definitionKey=leads answers fields as [{name, label, group}] — the recipient's names, the board's identity fields (lead.name, lead.email…) and every column under its slugged label ("Requested Amount" is requested_amount) — and syntax, the grammar: {{first_name}} for a field, {{first_name|lead.name|"there"}} for an either/or that tries each step and ends on the quoted words. Identity numbers and credit facts are never offered. It also answers addressColumns: the board's phone and email columns — and the plain text or number columns whose label says they hold one, as an imported board's do — as [{id, label, kind, default}], which is what addressColumnId on post_blasts_id_audience_board may name. Read it before writing a blast's words so every field named is one the audience can fill. GET /v1/blasts/merge-fields (access: workspace)

get_blasts_stats

ChatGPT
Read blast statistics How many blasts the workspace has sent, by status (sent, partial, failed, sending), for the Outreach page's header chips. Reach for it to answer "how is our outreach going" before opening any one blast. GET /v1/blasts/stats (access: workspace)

get_boards

ChatGPT
List the custom boards Every board the person may see — a table with its own columns and rows — with its access and view. rows=false answers each board's SHAPE: its columns with rows: [] and rowCount, how many rows the person may see — send it unless you need every row of every board, which is seconds on a deep workspace. summary=true answers those shapes smallest, each column as its id, label, kind and optionLabels. A board is found by its name in this list — its rows need not mention it. Each board carries its definitionKey (board-<hex>): a board is one door, so its records are read and worked through the definitions tools by that key, as a page's are by its own. The record pages (leads, deals…) are get_definitions; for every board and page, read both. GET /v1/boards (access: workspace)

get_boards_id

ChatGPT
Read one board, whole One custom board by id: its columns (each select column's choices with them), the layout it opens in, how far the CALLER may go on it, and EVERY ROW on it — unpaged, cells keyed by column id. A deep board is a deep answer (seconds on a board of tens of thousands of rows), so send rows=false when you need the board's SHAPE: the same board with rows: [] and rowCount, how many rows the caller may see. Leave it off only when you need every row at once; reach for get_boards_id_rows when you want a page of its rows narrowed by a search, ticked values or a sort, and for get_search when you only want to find one record by name. A pack page (leads, deals, contacts) is get_definitions_key for the shape and get_definitions_key_rows for one page of its rows. GET /v1/boards/:id (access: board:read)

get_boards_id_actions

ChatGPT
The buttons on a board's action bar The buttons on one custom board's action bar, in the order they are drawn: label, icon, kind (add_row, builtin or workflow), config, enabled flag and position. Any board reader may read it — the bar has to render. The pack-page twin is get_definitions_key_actions. GET /v1/boards/:id/actions (access: board:read)

get_boards_id_column-policies

ChatGPT
Which columns are locked on a board Every LOCKED column on one custom board, with the audiences each lock names — who may SEE the column's values and who may CHANGE them. Columns with no policy are simply absent. Workspace admin as well as manage on the board, because an audience says who in the workspace. The pack-page twin is get_definitions_key_column-policies. GET /v1/boards/:id/column-policies (access: workspace:admin)

get_boards_id_column-policies_effective

ChatGPT
What this caller gets on a board The CALLER's own two lists for one custom board: the columns whose cells will be stripped out of the rows they are served, and the columns their writes will be refused on. Any board reader may ask, and asking once is how a grid greys a cell out instead of drawing an empty one. It is an ANSWER, never the enforcement — the stripping and the refusals happen regardless. GET /v1/boards/:id/column-policies/effective (access: board:read)

get_boards_id_grants

ChatGPT
Who else can reach this board The colleagues one custom board has been shared with and the level each holds — read, write or manage. Manage on the board. A pack page's grants are its own (definitions/:key/grants). GET /v1/boards/:id/grants (access: board:manage)

get_boards_id_imports

ChatGPT
Board import history The board's past imports, newest first (limit defaults to 10, ceiling 50). Each run says what it was called, when it ran, how many rows landed and failed, the per-row results, whether it has been rolled back, and standing — how many of the rows it created are still on the board right now. Read this to find the runId an undo needs. GET /v1/boards/:id/imports (access: board:read)

get_boards_id_rows

ChatGPT
Read a page of a board's rows One page of a custom board's rows, narrowed by the service exactly as get_definitions_key_rows narrows a pack page: q searches every cell, filter[<column label or id>] (repeatable) keeps rows whose cell is one of the values (filter[$messages]=replied|sent keeps rows by who wrote last on their email and text threads; filter[$created]=>=2026-10-01 or $updated by when the row was made or last changed — a day, a month or an instant after >=, >, <= or <, repeated for a range, in UTC), sort + dir order by a stored column or $created / $updated, limit (default 200 when narrowed, at most 1000) and offset page. Answers {data: rows, page: {total, limit, offset}} where total is the whole narrowing's count. A filter or sort on a formula or lookup column is refused — read the board whole with get_boards_id. Read access on the board; a member on the assigned rung sees only the rows filed on them. fields names the only columns each row answers — column ids or labels, repeated or comma-separated (fields=Name,Status) — so reading many rows to export, count or summarise them carries the columns that matter instead of every cell; a field the board has no column for is refused. labels=true (a tool call's default; labels=false for ids) keys cells by column name and decodes relation/file cells. A DATE or datetime column takes the same bounds: ?filter[<its column id>]=>=2026-09-20 keeps the rows whose date is on or after the 20th (repeat for a range; a bare value is still the exact stored text) — so "deals funded in the last 30 days, and for how much" is ONE aggregates read over the Funded Date column, never the accounting ledger. dealRowId=<deal row id> keeps the rows linked to that deal on a page that links to Deals (one deal-link column), as on get_definitions_key_rows. GET /v1/boards/:id/rows (access: board:read)

get_boards_id_rows_rowId_files

ChatGPT
Files pinned to one board record The documents pinned to ONE record on a custom board — id, name, content type, size and document type — and what a file cell's picker offers there. Metadata only, so it answers even before file storage is configured. The pack-page twin is get_definitions_key_rows_rowId_files. Read access on the board. GET /v1/boards/:id/rows/:rowId/files (access: board:read)

get_boards_id_views

ChatGPT
Saved views on a board Every saved view on one custom board — the workspace's shared ones and the caller's own — each with its filter, sort and visible column ids. A pack page has no list route of its own: its views come inside get_definitions_key. Read access on the board. GET /v1/boards/:id/views (access: board:read)

get_boards_id_workspace-access

ChatGPT
Read a board's workspace access The level every seat with no grant of its own gets on a custom board: manage, write, read, or assigned (the default — a member sees only the rows whose assignee cell names them). Owners, admins, the board's maker and anyone with a grant are unaffected; below Enterprise every seat reads and writes regardless. Any reader of the board. GET /v1/boards/:id/workspace-access (access: board:read)

get_business_research

ChatGPT
Gets one business research record by its own ID. Prefer get_facility_business_research when you have an application or facility; use this only when you already hold a research record ID. It is read-only.

get_calendar_events

ChatGPT
Side effects: A calendar window read reconciles canceled meetings, updates mirrored tasks, and sends due reminders. List calendar events and reconcile meeting changes The caller's own calendar between two instants (from, to - RFC3339), read live from whichever vendor they connected (Google or Microsoft); entries this service booked carry \ours\ with the record they are about. This read also rings due meeting reminders and clears meetings deleted at the vendor, so it is not a cheap poll. Pass definitionKey instead to list RECORDS' meetings out of our own table, which reaches no vendor at all: with one rowId for one record, rowId repeated for several, or no rowId for every record on the page — \"meetings on all my leads\" or \"which of these have a demo booked\" is ONE call, never one per record; each meeting carries its rowId. account (an address) reads that one connected calendar; without it the read goes through the first that works and says so in accounts — every connected calendar with its state: read, ready, reconnect (its grant was refused: ask the person to reconnect it in Settings › Account) or no_calendar_scope. When a meeting is not where you expected, read accounts before saying there is none. GET /v1/calendar/events (access: workspace)

get_calendar_meet-sync

ChatGPT
Inspect Google Meet task sync Returns the integration administrator's enrolled Google connections and transcript document states, including generated task IDs, meeting IDs, and actionable review reasons. Use to verify whether a meeting transcript has been processed or needs its primary record linked. Does not fetch or change Google data. GET /v1/calendar/meet-sync (access: workspace)

get_capital-sources_rowId_lenders

ChatGPT
Read the lenders a capital source backs The lenders funded by one capital source. A capital source is the funder or investor BEHIND a lender, not a lender itself; the relation lives on the lender, so this answers what points at this row. GET /v1/capital-sources/:rowId/lenders (access: workspace)

get_cash-flows

ChatGPT
List cash flows The money expected to move on a deal and whether it actually did: per flow, the type (payment, disbursement, commission, fee, refund, adjustment), the direction (inbound or outbound), the method, the expected amount and due date, its place in the run, what actually landed and when, and its status — expected, pending, completed, failed, missed, cancelled or reversed. Filters: ?facilityId=, ?dealRowId= (a deal record on our own board), ?scheduleId=, ?status=, ?direction=, ?limit=. Cash flows are ALSO one of the four money boards: to show them as a page of rows with the app's own columns, saved views and facet filters, read get_definitions_key_rows with the key cash-flows. Come here for the ledger read and for the flow's own lifecycle verbs. GET /v1/cash-flows (access: workspace)

get_cash-flows_id

ChatGPT
Read a cash flow One expected movement of money: its schedule and deal, type, direction, method and currency, the expected amount and due date, its place in the run, what actually landed and when, its external transaction and batch ids, and its status with a failure's reason and code. GET /v1/cash-flows/:id (access: workspace)

get_cloud-files_source_id_content

ChatGPT
Read a picked cloud file's bytes One file from the caller's own connected storage, base64: :source is drive (a Google Drive file that is NOT a Google-native document - an .xlsx, .csv, .pdf or image) or onedrive (an id from get_onedrive_files). Answers {name, contentType, contentB64}. A Google Sheet is refused with 422 because it has no bytes of its own - read it through the Drive export instead. Refuses over 25 MB with 413 rather than truncating. Reach for it once a person has picked the file to attach. GET /v1/cloud-files/:source/:id/content (access: workspace)

get_comms_calls

ChatGPT
Read the call log Every call this workspace logged, newest first, across every thread — including threads filed under no record. Narrow with direction, or by record: definitionKey (the page's key, like "leads") or definitionId (its uuid), with rowId repeated for several records (rowId=a&rowId=b) or left out for every record on the page. ONE CALL FOR MANY RECORDS — never one call per record (2026-10-08: a review of 30 leads read their calls one at a time and ran out of time). Start here for "who called them, when, and how long did it go". GET /v1/comms/calls (access: workspace)

get_comms_connections_id_aliases

ChatGPT
Read a connected account's aliases The Gmail send-as aliases this workspace imported from one of the caller's OWN connected Google accounts, named by the connection's id (get_connections lists them). Each alias is an inbox of the caller's: get_comms_inboxes shows who is on it and get_comms_signatures how it signs. Somebody else's connection is a not-found. GET /v1/comms/connections/:id/aliases (access: workspace)

get_comms_connections_id_contacts

ChatGPT
Read a connected account's contact import How the last import of one of the caller's OWN connected Google or Microsoft accounts' saved contacts onto the Contacts board went, by the connection's id (get_connections lists them): status never, running, imported, failed (with the provider's words) or needs_consent (the account was connected before Specter asked for contacts; reconnect it), plus how many addresses were read, created and updated and when it last ran. Somebody else's connection is a not-found. GET /v1/comms/connections/:id/contacts (access: workspace)

get_comms_conversations

ChatGPT
List conversations The inbox: every conversation the person may see, with its participants and latest activity. Start here for anything about messages, replies or who last wrote. Narrow by record: name the page with definitionKey (its key, like "leads") or definitionId (its uuid), then rowId for one record, rowId repeated for several, or no rowId for EVERY record on the page — "all my leads", "the whole board", "these five records" are ONE call, never one per record; each conversation's records says which row it belongs to. A rowId with no page is refused. Add messages=N (up to 20) to get each conversation's newest N messages, oldest first, in the same answer — read what was said across a whole board without opening each thread. lastWord keeps the conversations whose newest message is from one side — lastWord=replied (they wrote last) or lastWord=sent (we did) — and a repeated bound on when it was written, in the rows route's $created shapes: lastWord=sent&lastWord=<2026-10-15 is every thread we wrote to last before the 15th that nobody answered. summary=true caps each conversation's records at five, with record_count saying how many it has. GET /v1/comms/conversations (access: workspace)

get_comms_conversations_id

ChatGPT
Read one conversation One conversation by its id, exactly as the inbox lists it: the other side (contact_identifier, contact_name), our line it is on (account_identifier), the medium, the records it is filed under, whether it is unread and its last word. Reach for this to know who a thread is with and which line answers it; its messages are get_comms_conversations_id_messages. Another workspace's conversation, or a private one the caller is not on, is a 404. GET /v1/comms/conversations/:id (access: workspace)

get_comms_conversations_id_messages

ChatGPT
Read a conversation's messages The messages of one conversation, in order, both directions, with context: the linked records and their facts, board_goal and customer_goal, and campaigns — the blasts whose copies were sent in this thread, each with the message this contact received, the goal its author set and whether they have replied. Read this before replying to or summarising a thread; when the contact is answering a campaign, its goal says what the reply should move them toward. board_goal is the linked board's or page's purpose (patch_boards_id sets a board's) and customer_goal the record's own column labelled Goal, Neo goal, Customer goal or Relationship goal. The most specific goal leads: the customer's, then the campaign's being answered, then the board's. GET /v1/comms/conversations/:id/messages (access: workspace)

get_comms_conversations_id_messages_messageId_opens

ChatGPT
Read estimated email opens Returns observed image-load statistics for one message in an accessible conversation. Only sent emails with tracking enabled are counted. Copies and forwarded messages share one signal; no recipient identity or proof of reading is inferred. GET /v1/comms/conversations/:id/messages/:messageId/opens (access: workspace)

get_comms_conversations_id_participants

ChatGPT
Read who is on a conversation Who is on this conversation — the colleagues on it and the outside contacts, each with their role. Read this before adding somebody, or to answer "who is on this thread". GET /v1/comms/conversations/:id/participants (access: workspace)

get_comms_inboxes

ChatGPT
Read the workspace's inboxes Every number and mailbox this workspace speaks through, each with the people staffing it. Start here for "which inboxes do we have", for who can read a line's messages, and for the keys the visibility settings are written with. GET /v1/comms/inboxes (access: workspace)

get_comms_labels

ChatGPT
List inbox labels The workspace's own inbox labels — words people made and filed threads under — each with id, name and tone. These are the made half of the Inbox's labels; the other half is what a thread is connected to, read from the records each listed conversation is filed under. Filter the conversations list to one label with label=<id>. GET /v1/comms/labels (access: workspace)

get_comms_messages_id_thread

ChatGPT
Read a message and its replies One message with the replies nested under it, for reading a single branch of a conversation without pulling the whole thread. GET /v1/comms/messages/:id/thread (access: workspace)

get_comms_numbers

ChatGPT
Side effects: Refreshes the caller’s stored Zoom line inventory when connected, updating line properties and removing lines no longer assigned by Zoom. Read the workspace's phone numbers Every line this workspace has, bought through us or assigned in Zoom, with what each one can carry, whether we still hold it, and which one is the sending number. Start here for "what numbers do we have". Its sending-number flag is only the workspace's DEFAULT line; which lines this person may use is get_comms_senders. GET /v1/comms/numbers (access: workspace)

get_comms_numbers_available

ChatGPT
Search phone numbers for sale Local US numbers the carrier has to sell in one area code, each with its SMS, MMS and voice capability. Nothing is stored: this is the menu a purchase is chosen from. GET /v1/comms/numbers/available (access: workspace:admin)

get_comms_opt-outs

ChatGPT
Read the do-not-contact list The addresses this workspace may not text, email or call, newest first: each phone number (E.164) or email address (lower-case), how it got there (inbound_reply when the contact replied STOP or similar, manual when a member listed it or moved the record to do not contact), the reason, the record it was listed against and who listed it. q matches the address, a number in any spelling, or the record name; limit and offset page it and total is the matched count. Read it before reaching out to somebody who may have asked to stop: every send to a listed address is refused. GET /v1/comms/opt-outs (access: workspace)

get_comms_recordings_id

ChatGPT
Read one call recording One recording: the file to play, the provider transcript, and whatever summary has been written. Never generates a summary — ask for one with the summarise route. GET /v1/comms/recordings/:id (access: workspace)

get_comms_records_definitionId_rowId_contacts

ChatGPT
Read a record's contact addresses The phone numbers and email addresses this record has actually been spoken to at, read from the threads themselves rather than from a contact table. Start here for "what number do we text them on" or "which address have we been emailing". GET /v1/comms/records/:definitionId/:rowId/contacts (access: workspace)

get_comms_records_definitionId_rowId_log

ChatGPT
Read a record's comms log Everything said with one record, in one answer: its calls, its texts (SMS and iMessage), its call recordings and its conversations, each newest first (calls, texts and conversations up to 200 deep; recordings all) — the same rows get_comms_calls, get_comms_texts, the recordings read and get_comms_conversations answer for that record, and a thread this person may not see is absent from all four. Start here for "what has happened with this deal on the phone and by message"; reach for the single reads to page further back or to read a whole board. GET /v1/comms/records/:definitionId/:rowId/log (access: workspace)

get_comms_records_definitionId_rowId_recordings

ChatGPT
Read a records call recordings Every recorded call filed under one record, newest first, each with its audio file, the transcript and the stored summary. Start here for "what was said on the calls about this deal". GET /v1/comms/records/:definitionId/:rowId/recordings (access: workspace)

get_comms_senders

ChatGPT
List the lines I can send from The mailboxes and texting numbers this seat may send a new message from: every inbox open to the workspace, every private one the seat is on, the seat's own connected email accounts, and every line when the seat is an admin or owner. Each line says whose it is: own is this seat's own line (its connected mailbox, its Zoom line), ownerEmail the colleague whose own line it is, and isPrimary only the workspace's default — never the seat's. Pass one as account_identifier on POST /v1/comms/send; a line not on this list is refused there. The same identifiers are what a blast's sendFrom takes (post_blasts, patch_blasts_id) to leave on one of these lines rather than the workspace's default. This list is the whole answer to what the seat may send from: get_comms_numbers' sending-number flag is only the workspace's default text line, and get_mailboxes only its watched inboxes. GET /v1/comms/senders (access: workspace)

get_comms_settings_visibility

ChatGPT
Read who may see messages by default The workspace's default visibility, keyed by medium ("sms") and by inbox ("sms:+15550100", "email:jordan@acme.com") — what becomes of a thread when nobody says otherwise. Any seat: what happens to a message you send is not a secret from you. GET /v1/comms/settings/visibility (access: workspace)

get_comms_signatures

ChatGPT
Read the email signatures Every signature the workspace has, one per sending address, with the text appended to mail leaving that address. Lists them all, because a signature is published text a stranger reads. Read it before ending an email with a sign-off: the email is signed for its sending address already. A text is signed by nothing here, so its sign-off is the person's. GET /v1/comms/signatures (access: workspace)

get_comms_texts

ChatGPT
Read the text message log Every SMS and iMessage this workspace sent or received, newest first, across every thread. Narrow with direction, or by record: definitionKey (the page's key, like "leads") or definitionId (its uuid), with rowId repeated for several records (rowId=a&rowId=b) or left out for every record on the page. ONE CALL FOR MANY RECORDS — never one call per record (2026-10-08: a review of 30 leads read their texts one at a time and ran out of time). Start here for "what have we texted them". GET /v1/comms/texts (access: workspace)

get_communication_channel_for_entity

ChatGPT
Gets the communication channel for a specific entity visible to the authenticated user. Use before reading or sending messages tied to a facility, submission, lead, counterparty, or lender. A successful result exposes channel.id/channel_id in the model-visible response; pass that exact value to send_channel_message. If channel is null, no channel exists yet — open one with create_communication_channel and use the id it returns; do not guess REST paths. It is read-only.

get_connections

ChatGPT
List connected accounts List the caller's provider accounts authorized for the current workspace only. Connecting an account in another workspace does not authorize it here. GET /v1/connections (access: workspace)

get_connections_id_mail-folders

ChatGPT
List connected mailbox folders Read or choose native Yahoo/AOL folders for your own connected workspace mailbox. Requires the current mailbox owner and active consent. Folder names and derived data are prohibited from advertising use. GET /v1/connections/:id/mail-folders (access: workspace)

get_credit-memos

ChatGPT
List credit memos The underwriting memos written on deals — the write-up that recommends a decision, with the approval trail on it. Each carries its type (initial, renewal, modification, exception), title, status (draft, in_review, approved, rejected, expired), the amount requested and the amount recommended, the recommended rate, term, product and decision (none, approved, declined, counter_offer) with its conditions, and who reviewed or signed it off. Filters: ?facilityId=, ?status= (?status=in_review is the sign-off queue), ?limit=. GET /v1/credit-memos (access: workspace)

get_credit-memos_id

ChatGPT
Read a credit memo One credit memo in full: its nine narrative panes — executive summary, business overview, financial analysis, industry analysis, risk assessment, collateral analysis, ownership analysis, strengths and weaknesses, recommendation — beside the amounts and terms it recommends, the conditions on them, its status, and the review, approval or rejection with who signed it and when. GET /v1/credit-memos/:id (access: workspace)

get_credit_report

ChatGPT
Gets one stored credit report, optionally with its tradelines in the same call. Use when the user asks about a specific credit pull, its scores, or the individual tradelines behind it. Reads a stored pull only — it does not order a new one. It is read-only.

get_definitions

ChatGPT
Side effects: For admins, an empty workspace installs its default pack; otherwise this read can reconcile pack columns, choices, and lifecycle data after answering. List record pages and synchronize installed page schemas Every record page in the workspace — leads, deals, contacts and any custom page — with its key, statuses and views. Pass summary=true to list the pages without their columns; read ONE page's column ids with get_definitions_key before listing or changing its rows. GET /v1/definitions (access: workspace)

get_definitions_key

ChatGPT
Read a page's definition: its columns, statuses and roles One record page by key, with its columns (ids, kinds, options), statuses and saved views. The schema a row's cells are keyed by. GET /v1/definitions/:key (access: workspace)

get_definitions_key_actions

ChatGPT
List a page's row actions The buttons on one record page's action bar, in the order they are drawn: each with its label, icon, kind, config, enabled flag and position. Any workspace seat may read it — the bar has to render — while adding or reshaping one is an admin's. Read this before patching or reordering an action, to learn the ids. GET /v1/definitions/:key/actions (access: def:read)

get_definitions_key_aggregates

ChatGPT
Sum, count or average a page's records The calculations strip of a record page: how many rows, and one figure per column that carries a calculation (sum, average, median, min, max, filled, empty, percent filled/empty, unique, earliest, latest) — the way to answer "what is the total loan amount" or "how many leads have a phone" without listing rows. Takes the rows route's narrowing (?q=, ?filter[<column label or id>]=<value> repeated, ?group=<groupId>, ?view=<tab>, and ?dealRowId= on a page that links to Deals) and figures over the rows in it. Computed columns (formula, lookup) are left out. ?groupBy=<column label or id> (twice for a cross-tab; a date as <column>:month|week|day) answers the count and figures per group, largest first — "submissions by lender and status" in one call; a board is board-<id without dashes>. ?filter[$created]=>=2026-10-01 narrows by when the row was made ($updated: last changed) and ?groupBy=$created:month groups by it, so "deals by stage this month" needs no date column. A DATE or datetime column takes the same bounds: ?filter[<its column id>]=>=2026-09-20 keeps the rows whose date is on or after the 20th (repeat for a range; a bare value is still the exact stored text) — so "deals funded in the last 30 days, and for how much" is ONE aggregates read over the Funded Date column, never the accounting ledger. ?calculate=Funded:sum (repeated; any of sum, average, median, min, max, filled, empty, percent_filled, percent_empty, unique, earliest, latest) figures a column the page carries no calculation on, or replaces the one it carries — the strip otherwise totals only the columns its admin set, so "for how much" is asked here. ?view=<tab> reads one of the page's tabs (a band of its status — Underwriting on deals — or one status), as the rows read does; a status filter valued by a tab's name means that tab. GET /v1/definitions/:key/aggregates (access: def:read)

get_definitions_key_child-pages

ChatGPT
List the child pages a page accepts Which other pages may hang subrecords under this one, what each section is called, and the order the sections are drawn in. Read this before offering a subrecord picker: a page that accepts nothing draws no Subrecords section. GET /v1/definitions/:key/child-pages (access: def:read)

get_definitions_key_column-policies

ChatGPT
Read a page's column locks Every LOCKED column on one record page, with the audiences each lock names — who may see the column and who may write it. Columns with no policy are simply absent. Workspace admin, because an audience says who in the workspace may see what; a member asking what applies to THEM reads get_definitions_key_column-policies_effective instead. GET /v1/definitions/:key/column-policies (access: def:admin)

get_definitions_key_column-policies_effective

ChatGPT
Read the column locks in force for the caller What the CALLER's own seat gets on one record page: the columns whose cells will be stripped out of their rows, and the columns their writes will be refused on. Any seat may ask, and it is cheap enough for a board to ask on every load instead of guessing per cell. It is an ANSWER, not the enforcement — a client that never calls it still receives stripped rows and still has those writes refused. GET /v1/definitions/:key/column-policies/effective (access: def:read)

get_definitions_key_entity-groups

ChatGPT
List relationships on this page Every entity group on the page: several rows that are ONE commercial relationship (a parent and its subsidiaries, an operator and its franchises), with each group's kind, principal and member count. Reach for it when the question is 'which of these companies are the same party'. NOT the named record groups (…/groups): those are curated collections a row may be in several of; a relationship is exclusive. GET /v1/definitions/:key/entity-groups (access: def:read)

get_definitions_key_entity-groups_groupId_members

ChatGPT
Read a relationship's members Who is in one entity group, oldest first, each with the role it plays inside the relationship and whether it is the principal. GET /v1/definitions/:key/entity-groups/:groupId/members (access: def:read)

get_definitions_key_facets

ChatGPT
Read a page's filter facets Per-column value counts for a record page — how many rows sit in each status, stage or option. The cheap way to answer "how many … are …" without listing rows. Takes the rows route's narrowing — ?q= (substring over every cell), ?filter[<column label or id>]=<value> repeated, ?group=<groupId>, ?dealRowId= on a page that links to Deals — and then counts only the rows in it, so a paged grid can draw its filter panel without the rows; ?filter[$created]=>=2026-10-01 (or $updated) counts only the rows made (last changed) in that span. Also answers a $messages column (Messages: replied / sent) counting records by who wrote last on their email and text threads — the cheap way to answer "how many leads replied"; filter the rows route with the same word to list them. A relation column is counted by the linked record (its 50 largest), labelled with the record's name. A DATE or datetime column takes the same bounds: ?filter[<its column id>]=>=2026-09-20 keeps the rows whose date is on or after the 20th (repeat for a range; a bare value is still the exact stored text) — so "deals funded in the last 30 days, and for how much" is ONE aggregates read over the Funded Date column, never the accounting ledger. ?view=<tab> reads one of the page's tabs (a band of its status — Underwriting on deals — or one status), as the rows read does; a status filter valued by a tab's name means that tab. GET /v1/definitions/:key/facets (access: def:read)

get_definitions_key_grants

ChatGPT
Who is held back on a pack page The members an admin has held to none (kept off), read or write on one pack page — Leads, Deals, Merchants. Anyone not listed reads and writes as every seat does; owners and admins are never held back. Any seat may read it. GET /v1/definitions/:key/grants (access: def:read)

get_definitions_key_groups

ChatGPT
List a page's record groups The named record groups on one page — the collections a person makes and files records into — each with its description, colour, status-group filing, position and how many records it holds. Groups are row DATA, so any seat may read and make them, unlike the page's status vocabulary. A group is not a status: a record can be in several of them, or in none. GET /v1/definitions/:key/groups (access: def:read)

get_definitions_key_groups_groupId_rows

ChatGPT
List the records in a group The records in one group, in the board's own order, each as cells keyed by column id. A group is a view over the page, so this reads no more of the page than the page itself would: the caller's column policy and row wall apply here too. A group id that is not on this page is a 404. GET /v1/definitions/:key/groups/:groupId/rows (access: def:read)

get_definitions_key_import_jobs_jobId

ChatGPT
Side effects: Advances an import by one chunk, potentially updating existing records. Process the next import batch and report progress How far one import job has got: state (running | completed | failed), total rows it was given, done rows it has finished with, failed rows refused, imported rows landed, and the first 20 refusals with their row numbers in the file. This read is what MAKES the progress — it lands the job's next chunk of 200 rows before it answers, because there is no background worker — so poll it about once a second until the state is terminal, and a 5000 row file takes about 25 reads. Two reads at once never land the same chunk twice; the second answers the counts as they stand. done counts every finished row whatever became of it, so it reaches total even on a file with duplicates the policy skipped. A job nobody reads for 15 minutes is settled failed with "the import stopped being polled". The full per-row report and the undo live on get_definitions_key_imports. GET /v1/definitions/:key/import/jobs/:jobId (access: def:read)

get_definitions_key_imports

ChatGPT
Side effects: Advances the oldest import job and can mark an abandoned job failed. Advance pending imports and list import history One page's import history, newest first: each run's file, counts, error and per-row results, plus how many of the rows it made are STILL on the page — which is what says whether an undo is worth anything. limit defaults to 10 and is capped at 50. This read also ADVANCES the page's oldest running import job by a chunk, because there is no background worker: a job whose wizard was closed moves on when somebody comes looking for it, and one abandoned long enough is settled failed here. GET /v1/definitions/:key/imports (access: def:read)

get_definitions_key_labels

ChatGPT
List a page's record names Every record on a page this caller may see, newest first, as its id and the text of ONE column (?column=<columnId>, usually the page's first column — its name). The light read for resolving or offering records by name: use it instead of reading the whole page when only ids and names are needed. A column hidden from this caller answers blank names. GET /v1/definitions/:key/labels (access: def:read)

get_definitions_key_replies

ChatGPT
Who wrote last on each record The reply marks of a record page: for every record with an email or text thread you may read, who wrote last — direction inbound (their message is the newest, so it is unanswered) or outbound (ours is, and they have not answered yet) — with the medium (email, sms or imessage), from (the contact, or the workspace member who sent ours), the email's subject, a one-line preview, at, the thread's conversationId and count (how many in a row from that side). Workspace-wide: a reply from anyone in the workspace makes a record outbound. A private thread counts only for the seats that may read it; calls and voicemail are not counted; records nobody has written to are absent. With nothing else it is every mark on the page at once, for drawing a board. It takes the rows route's narrowing — ?q=, ?filter[<column label or id>]=<value> repeated ($messages and its bound included), ?group=<groupId>, ?rowId=<id> repeated — and ?limit= / ?offset= (page.total says how many answer), so "who hasn't answered in 4 days" is ONE read: ?filter[$messages]=sent&filter[$messages]=<2026-10-15 (today minus four days, from the clock) names each of them with the conversationId to follow up on and the message they did not answer. ONLY THE NEWEST MESSAGE PER THREAD: an earlier reply that was since answered does not show here, so "no replies" is the wrong reading of an outbound mark. To see every reply, read the threads with their messages — get_comms_conversations with definitionKey (and rowId repeated, or none for the whole page) and messages=N — and look at the inbound ones (2026-10-07: Josh saw replies Neo said were not there). A date column takes bounds in its filter too (filter[<its id>]=>=2026-09-20), as on the rows read. ?view=<tab> reads one of the page's tabs (a band of its status — Underwriting on deals — or one status), as the rows read does; a status filter valued by a tab's name means that tab. dealRowId=<deal row id> keeps the rows linked to that deal on a page that links to Deals (one deal-link column), as on get_definitions_key_rows. GET /v1/definitions/:key/replies (access: def:read)

get_definitions_key_rows

ChatGPT
List or search a page's records The rows of one record page, each as its cells plus its position and source. How to list leads, deals or contacts — and a board you made, by its board-<hex> key (its definitionKey): one door for both. Answers the LIVE board; pass ?archived=true for the archived records instead, most recently archived first. A row carries archived_at only while it is archived. The four MONEY boards — payments, payment-schedules, cash-flows, obligations — are read here too and are the ledger itself: their rows carry source ledger, the Deal column is a relation cell naming the deal record, and they take ?view= (one of that board's own view keys) and ?dealRowId= (one deal's rows, which is what the deal's Servicing tab shows). On any other page ?view= names one of its TABS — a band of its status (the deals page's Underwriting, a board's banded status option) — or one status, by name, and answers the rows in it; a word that is neither is refused naming the tabs. On any other page that links to Deals — Submissions, Offers, Contacts, a board with a Deal column — ?dealRowId=<deal row id> is the filter on that one link column: that deal's records, so "any offers yet on Vireo" is offers/rows?dealRowId=<Vireo's id>. A page with no column linking to Deals, or more than one, is refused naming them and the deal's own reads (deals/rows/<id>/detail, …/offers, …/links). A saved view is sent as what it holds: ?filter[<column label or id>]=, ?sort= and ?q=, read from the page's definition. ?limit= and ?offset= page any board, and page in the answer says how many rows the narrowing holds. ?rowId=<id> repeated reads exactly those records (up to 100), each as the one-record read answers it, in the order sent — a record's related submissions, offers or contacts are ONE read, never one call per id; it takes labels, fields and expand (detail, or context on any page) and refuses the page's narrowing. A paged read of an ordinary board also takes ?q= (case-blind substring over every cell), ?filter[<column label or id>]=<value> repeated (exact match on the stored cell; an empty value matches a blank cell), ?sort=<column label or id>&dir=asc|desc (numbers as numbers, dates as ISO text, else as text; empties last) and ?group=<groupId> (record-group membership); a computed column (formula, lookup, breakdown, files, documents) filters and sorts like any other, read over at most 5,000 rows the other filters leave (more is refused, 413, naming how to narrow); a documents column (a record's document checklist, e.g. Deals' Documents) takes missing, requested or complete — ?filter[Documents]=complete is everyone not missing docs. Any of the four without limit answers the first 50 rows (page.total says how many match). ?filter[$messages]=replied|sent narrows by who had the last word (get_definitions_key_replies says what each means), and a bound beside it by when that last word was written, in $created's shapes: ?filter[$messages]=sent&filter[$messages]=<2026-10-15 is every record whose last word is ours, written before the 15th and still unanswered — the date is today's minus N, never "4d". ?filter[$created]=>=2026-10-01 (or $updated) narrows by when the row was made (last changed) — a day, a month or an instant after >=, >, <= or <, repeated for a range, in UTC — and ?sort=$created orders by it. ?expand=context adds context keyed by row id — each record's open tasks, files, document requests, latest comms and timeline, and its links both ways (a deal's relations and lifecycle too), on ANY page, for up to 25 rows; reading rows by rowId, it is the door's default. On the deals board, ?expand=detail adds details keyed by row id — each deal's full record (lifecycle, owed amounts, files, document requests, tasks, activity) for up to 25 rows — so reviewing a page of deals is ONE read, never one detail call per deal. ?filter[$bank]=linked|error|none narrows by the record's Plaid bank links — linked connected, error the merchant must sign in …

get_definitions_key_rows_assignees_next

ChatGPT
Pick the next assignee in a rotation Of the accounts named in accounts, the one who has gone longest without role on this page. It reads the assignment trail the activity log already holds, so it stays right when somebody assigns a record by hand — the round-robin half of the shipped assign-new-records rule. GET /v1/definitions/:key/rows/assignees/next (access: def:read)

get_definitions_key_rows_due

ChatGPT
Leads due today The leads whose follow-up date has arrived and that are not currently snoozed, oldest first. This is the work the desk owes right now; a snoozed lead reappears here by itself the moment its snooze lapses. Read it before asking what to work on. GET /v1/definitions/:key/rows/due (access: def:read)

get_definitions_key_rows_facets

ChatGPT
Count a page's records by column value How many deals sit on each status and each stage — the counts behind the deals board's filter chips, as a column per facet with its values and totals. Empty values are left out and a facet with nothing in it is dropped, so every column that comes back has options. A column this caller's column policy hides has no facet at all. Deals page only. GET /v1/definitions/:key/rows/facets (access: def:read)

get_definitions_key_rows_rowId

ChatGPT
Read one record One row of a record page, whole: every cell, and next_statuses — the statuses it may move to from where it stands, the lifecycle's own on a leads page. Read this before changing or describing a specific record; several records of one page are ONE rows read with rowId repeated, never a call each. labels=true (a tool call's default; labels=false for ids) keys cells by column name and decodes relation/file cells. expand=context answers beside it what hangs off the record, keyed by its id: its open tasks (the done ones counted), files, document requests, latest emails, texts and calls (comms), latest timeline (recent) and its links both ways (links: what it points at, what points at it, its subrecords and parents, each named with its status), and on a deal its relations by name, lifecycle, risk signals and closings — what the record's tabs show, in one read. A board somebody made reads here too, by its definitionKey (get_boards answers it). Many at once: get_definitions_key_rows GET /v1/definitions/:key/rows/:rowId (access: def:read)

get_definitions_key_rows_rowId_actions

ChatGPT
List what can be done to a record Everything that can be done to ONE record, read off the catalog: every route this seat may call that takes this record, each with the record's own arguments filled in (args), whether Neo would ask the person first (asks), its permissions area, and tab_match for the ones the open pop-up tab is about (pass tab, e.g. emails or files — those come first). Beside them, the record at a glance: its label, status and next_statuses, its phone numbers and email addresses (cells first, then the ones its threads used), and how many conversations and files it has. Reach for it for "what can I do here"; to do one thing, find_tools with what it is. Not the page's configured row buttons (get_definitions_key_actions). approvalMode is the person's own (auto or review; review when left out). GET /v1/definitions/:key/rows/:rowId/actions (access: def:read)

get_definitions_key_rows_rowId_activity

ChatGPT
Read a record's activity feed One record's own timeline, newest first: who did what to it and when, with the before-and-after values on an edit. limit narrows it. Columns this caller's column policy hides are stripped out of the entries too, so a locked column's history is not readable through the feed. Any record page; for just the status ladder, get_definitions_key_rows_rowId_status-history is this same feed filtered. GET /v1/definitions/:key/rows/:rowId/activity (access: def:read)

get_definitions_key_rows_rowId_affiliate-invitations

ChatGPT
List a record's outside affiliates The outside affiliates invited onto ONE record of a pack board — a provider on a treatment row, co-counsel on a case — live AND dead, newest first, each with the role it was minted under. The record must be one the caller can see. GET /v1/definitions/:key/rows/:rowId/affiliate-invitations (access: def:read)

get_definitions_key_rows_rowId_ai_insights

ChatGPT
Read a deal's AI insights The AI summary and bullet insights stored on one deal, with the model that wrote them and when. There is at most one live read per deal; regenerating replaces it. Not found means nothing has been generated yet, which is a state rather than an error — never an empty summary, which the schema refuses to store. GET /v1/definitions/:key/rows/:rowId/ai/insights (access: def:read)

get_definitions_key_rows_rowId_api-connection

ChatGPT
Read a lender's API connection How this workspace reaches one Lenders record's submission API (SmartBiz, Bitty, PIRS, Spectra, Lendr, ECG, Credibly): provider, endpoint, enabled, and which credentials are on file — apiKeyConfigured with the key's last four, brokerIdConfigured, portalCredentialsConfigured. The credentials themselves are write-only and never answered. 404 when the lender has no connection; then a send to a lender whose record names a portal or API is refused lender_api_not_connected. Lenders pages only. GET /v1/definitions/:key/rows/:rowId/api-connection (access: def:read)

get_definitions_key_rows_rowId_assignments

ChatGPT
Read who is assigned to a record, by role Everyone working one record, grouped by role, the role's primary first — the SET behind the record's person cell. A role nobody holds is absent rather than present-and-empty, and an id that names no record on this page is a 404, never an empty set. Deals carry agent, closer, originating_agent and processing_agent; leads, submissions and merchants carry owner. GET /v1/definitions/:key/rows/:rowId/assignments (access: def:read)

get_definitions_key_rows_rowId_bank-statements

ChatGPT
List a record's bank statements One deal's bank statements as the processor read them: each file with the months extracted from it — revenue, deposits, average daily balance, NSFs, the positions found in it and the reasoning behind each figure — and the rollup those months add up to. get_definitions_key_rows_rowId_statements is the other half: the files and spans a person filed, without any of this arithmetic. Deals page only. GET /v1/definitions/:key/rows/:rowId/bank-statements (access: def:read)

get_definitions_key_rows_rowId_books

ChatGPT
A client's books One business's or household's own books, kept on its record: whether they are set up (ready), the chart of accounts (each with code, name, kind - asset, liability, equity, income, cogs, expense - and, for a linked bank account, bankAccountId; a bank linked since the last write reads with an id bank:<id> that any write accepts), the categorization rules (payee or bank-category, from the template or a person), how many synced bank lines wait for review, and how many bank accounts the books follow. The firm's OWN books are the accounting routes, never this. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. GET /v1/definitions/:key/rows/:rowId/books (access: def:read)

get_definitions_key_rows_rowId_books_entities

ChatGPT
Every report a client stands behind On a Clients record: its own household books and each Businesses record whose relation names this client, each with its period figures (income, net income, assets, liabilities, equity, the balance check, uncategorized count) or ready: false when its books are not set up. ?from=/?to= as the report. A business the seat cannot open is not listed. One customer to many financial reports, in one read. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. GET /v1/definitions/:key/rows/:rowId/books/entities (access: def:read)

get_definitions_key_rows_rowId_books_entries

ChatGPT
List a client's journal entries The record's journal, newest first: opening balances and manual entries (payroll, loans, depreciation, adjustments) with their lines, each a debit-positive amount in cents on one account. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. GET /v1/definitions/:key/rows/:rowId/books/entries (access: def:read)

get_definitions_key_rows_rowId_books_report

ChatGPT
A client's P&L and balance sheet The record's profit and loss over ?from= to ?to= (YYYY-MM-DD; from defaults to January 1 of to's year, to to today) and its balance sheet at the end of to, built from its categorized bank lines, opening balances and journal entries, on a cash basis in USD. Uncategorized lines report as 'Uncategorized income/expenses (to review)' and are counted in uncategorized; balanceSheet.check is 0 when the books balance. For every report a client stands behind, read books/entities on the client. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. GET /v1/definitions/:key/rows/:rowId/books/report (access: def:read)

get_definitions_key_rows_rowId_books_transactions

ChatGPT
List a client's categorized bank lines The record's synced, posted bank lines, newest first, each with where it is filed (accountId) and by whom (filedBy: person, rule, or empty while it waits for review). ?status=review keeps only what waits; ?from=/?to= (YYYY-MM-DD) narrow the days; ?limit= up to 500 (default 200). total and forReview count the whole narrowing. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. GET /v1/definitions/:key/rows/:rowId/books/transactions (access: def:read)

get_definitions_key_rows_rowId_closings

ChatGPT
How this deal ended Every time this deal was closed out, newest first: whether it was won, lost or declined, the status it landed on, the reason the desk gave, the closing note and the follow-up task. Empty for a deal still running. Reach for it to answer "why did we lose this" or "what was the reason we declined". GET /v1/definitions/:key/rows/:rowId/closings (access: def:read)

get_definitions_key_rows_rowId_condition-shares

ChatGPT
List a deal's lender checklist links One deal's read-only checklist links for its lender, newest first, LIVE AND DEAD, each {id, dealId, submissionShareId, createdBy, status (live, expired, revoked or withdrawn — the last when the submission it was sent with was withdrawn), createdAt, expiresAt, revokedAt}. The link's address is never here: it exists once, in the answer that minted it. Deals page only; a deal outside the caller's reach is a not-found. GET /v1/definitions/:key/rows/:rowId/condition-shares (access: def:read)

get_definitions_key_rows_rowId_contract_documents

ChatGPT
List a deal's contract documents The papers issued on one deal — rendered documents, each with the template version it was built from — newest first. The path says contract in the singular and carries no contract id because the DEAL ROW is the contract's address in this service. Deals page only. GET /v1/definitions/:key/rows/:rowId/contract/documents (access: def:read)

get_definitions_key_rows_rowId_contract_signature-requests

ChatGPT
List a deal's signature requests Every envelope this deal's contract has been through, newest first, with each signer's state: waiting, viewed, signed or declined. GET /v1/definitions/:key/rows/:rowId/contract/signature-requests (access: def:read)

get_definitions_key_rows_rowId_contract_signature-requests_requestId_document

ChatGPT
Read the signed contract The self-contained HTML artefact a completed envelope produced: the words as sent, every signature inline, and the audit trail. Not found while the envelope is still open. GET /v1/definitions/:key/rows/:rowId/contract/signature-requests/:requestId/document (access: def:read)

get_definitions_key_rows_rowId_credit-records_family

ChatGPT
Read a deal's credit records This deal's credit records of one family - credit_memo, decision or business_research - newest first. A decision carries what was approved, or the reason it was declined. GET /v1/definitions/:key/rows/:rowId/credit-records/:family (access: def:read)

get_definitions_key_rows_rowId_credit-reports

ChatGPT
List a record's credit reports One deal's credit pulls as the processor read them: each report with its bureau, type, subject, score, balances, limit, utilisation and inquiry count, and every tradeline under it. Deals page only; for a credit DECISION or a memo the desk wrote, read get_definitions_key_rows_rowId_credit-records_family instead. GET /v1/definitions/:key/rows/:rowId/credit-reports (access: def:read)

get_definitions_key_rows_rowId_debt-schedules

ChatGPT
List a record's debt schedules Every business debt schedule on one record — the outstanding positions a merchant is carrying, each with its revision and whether it has been issued. Reach for it to answer what a business already owes before pricing anything new. GET /v1/definitions/:key/rows/:rowId/debt-schedules (access: def:read)

get_definitions_key_rows_rowId_detail

ChatGPT
Read one record in full: cells, people, files, activity The deal record page's opening read, in one call: the deal's cells, its cross-board relations, both lifecycle tracks, its files, what is still owed on it, its tasks, its people by role and its latest activity. Reach for this rather than assembling a record out of six reads; reach for get_definitions_key_rows_rowId_relationships when only the links are wanted. summary=true names each related record as {id,label,status,statusLabel} instead of its whole row, and lists only the open tasks with closedTaskCount for the rest. Deals page only. ONE deal: for a page of deals at once, read the rows with expand=detail instead of calling this per deal. GET /v1/definitions/:key/rows/:rowId/detail (access: def:read)

get_definitions_key_rows_rowId_disclosure

ChatGPT
Read a submission's financing disclosure The live state-mandated commercial financing disclosure on one submission record, with the rendered words and how far along it is (generated, out for signature, signed). 404 means there is none, which is the ordinary state of a submission whose merchant is not in a state with a disclosure law. GET /v1/definitions/:key/rows/:rowId/disclosure (access: def:read)

get_definitions_key_rows_rowId_document-analyses

ChatGPT
List a record's document analyses Every analysis ever asked for on this record's documents, newest first, each with its attempt number, its status (queued, running, done, failed) and the reason it failed. Answers "has this statement been analysed yet" and "why is there no result for this document". Every attempt is listed, not only the latest per file. GET /v1/definitions/:key/rows/:rowId/document-analyses (access: def:read)

get_definitions_key_rows_rowId_document-facts

ChatGPT
Read what every file on a record says What's in a record's files, in one read: every live file pinned to the record, newest first, each with fileId, name, documentType ("" when unclassified) and documentTypeSetBy (processor, person or file_name), landed, attempt (the newest processing attempt on it in ANY job — outcome pending, started, succeeded, partial, failed or skipped, with reason; null when it was never sent), readByJobId ("" when nothing was read off it) and overview: summary (the processor's own one- or two-sentence description, "" when it wrote none), keyFacts [{label, field, value, page, sensitive}], parties, dates, amounts (as printed) and proposedType (what an unplaced document looks like, a proposal and never a type). Fields come from the newest job that read EACH file, not the record's newest job, so a file read in an earlier run still answers. A sensitive fact is a masked whole identifier ("••••1234"): the whole value is never in this read. Ask this first for "what's in this loan file" or "which document says X"; for every field of one file, get_files_id_reading. Files a private conversation brought stay their owner's; truncated is true past 400 files. Any seat that can open the row. GET /v1/definitions/:key/rows/:rowId/document-facts (access: def:read)

get_definitions_key_rows_rowId_document-forms

ChatGPT
List the document template forms a record fills The newest version of every PDF form in Settings › Document templates written against this record's page — a Tax Guard or IRS 8821, a retainer, any form a workspace uploaded once — on any board, a law case, a deal, a CRE loan or a made board alike. Each carries its pages and fields (the record word, the sender or the signing party that fills each). Fill one through post_definitions_key_rows_rowId_lender-forms_formId_preview and …_issue on this record. Read-only; the record's own read. GET /v1/definitions/:key/rows/:rowId/document-forms (access: def:read)

get_definitions_key_rows_rowId_document-requests

ChatGPT
List the documents requested from a record What is still owed on one record: every document asked for, open ones first and oldest ask first inside each group, with the period each covers. This is what the desk has ASKED for; get_definitions_key_rows_rowId_documents_checklist is the wider question of what the record's form asked for and what has actually arrived. GET /v1/definitions/:key/rows/:rowId/document-requests (access: def:read)

get_definitions_key_rows_rowId_document-review

ChatGPT
List a record's document values waiting for review The documents whose processed values are still held for a person on this record: one entry per file, oldest wait first, with its fileId, name and waiting — how many of its values wait (outcome review) on this record. Settled values (applied, accepted, rejected) are not counted, and neither are a form's held values (get_definitions_key_rows_rowId_intake-review). Each value is settled on its file: get_files_id_result-changes lists them and the accept/reject routes beside it decide one. Files a private conversation brought stay their owner's; a value in a column this caller's policy hides is left out. [] when nothing waits. Any seat that can open the row. GET /v1/definitions/:key/rows/:rowId/document-review (access: def:read)

get_definitions_key_rows_rowId_document-tabs

ChatGPT
Read a record's document tabs The record dropdown's document tabs, as the page's pack template declares them: one entry per tab in the declared order, each with its documentType, label and fields (field is the processor's name, label the words for a header), and files — every live file of that type on the record, newest first, with landed (the bytes are in storage) and values, the declared fields read off it by the newest processing job that read the file, verbatim as the engine reported them (a field it did not read is absent). Files a private conversation brought stay their owner's. A page whose template declares no tabs (every finance page, a made board) answers []. Any seat that can open the row. GET /v1/definitions/:key/rows/:rowId/document-tabs (access: def:read)

get_definitions_key_rows_rowId_documents_checklist

ChatGPT
Read a deal's document checklist What one record was ASKED for and where each paper stands — the expected set, what has arrived and what is missing. The expected set comes from the FORM the merchant filled in (a form's file fields are the documents it asks for, unioned across every submission filed against the record), never from a list of required documents this service does not have. A record no form created answers formBorn: false and lists only its open asks and the typed files already on it — nothing is missing when nobody asked. GET /v1/definitions/:key/rows/:rowId/documents/checklist (access: def:read)

get_definitions_key_rows_rowId_email-rules

ChatGPT
Read lender email processing rules Effective lender reply instructions and regex link-extraction rules, with platform defaults and explicit workspace overrides. Reads only an accessible record on a Lenders page. GET /v1/definitions/:key/rows/:rowId/email-rules (access: def:read)

get_definitions_key_rows_rowId_feed

ChatGPT
Read a record's merged feed One paged, time-ordered timeline for a record, merging four sources it already has: activity-log entries (which carry stage moves, signing and underwriting events), internal thread comments, texts/emails/calls linked to the record, and documents that landed. Each item carries an actor, a time, a one-line sentence, a link, its reactions and whether the caller follows the record. Reach for this to answer 'what has happened on this record' in full; the narrower /activity route answers only 'what did people do to it'. Page with the returned nextCursor; narrow with kind=activity|thread_message|comms|document. GET /v1/definitions/:key/rows/:rowId/feed (access: def:read)

get_definitions_key_rows_rowId_files

ChatGPT
List a record's files The files pinned to one record — name, content type, size, document type and any period they cover — oldest first, deleted ones left out. Metadata only, so it answers even before file storage is configured; a download URL is a separate signed read. Any record page. GET /v1/definitions/:key/rows/:rowId/files (access: def:read)

get_definitions_key_rows_rowId_financials

ChatGPT
Read a record's financials The deal Highlights' summary: monthly revenue, average daily balance, NSFs, positions and the score, each named with the job it came from. Computed from the same extracted months the statements tab shows, with the same arithmetic, so a headline cannot disagree with the table under it. A deal nothing has been read for answers empty job ids rather than a 404 — 'not processed' is a state. Deals page only. GET /v1/definitions/:key/rows/:rowId/financials (access: def:read)

get_definitions_key_rows_rowId_folders

ChatGPT
List a record's document folders The folders a record's Documents are filed in, in the order the desk put them in, each with how many files and sub-folders it holds. A folder with a null parentFolderId sits at the record's Documents root. GET /v1/definitions/:key/rows/:rowId/folders (access: def:read)

get_definitions_key_rows_rowId_health

ChatGPT
Read a record's health score One deal's health score, computed ON THIS READ from facts the service already holds — how long it has sat in its stage, what documents are outstanding, its submissions, and how recently anybody made contact — with the factors that produced the number. There is no stored column and no refresh route, deliberately: a caller who wants a fresh score calls this again. Deals page only. GET /v1/definitions/:key/rows/:rowId/health (access: def:read)

get_definitions_key_rows_rowId_heartbeat_actions

ChatGPT
List the heartbeat's proposed actions for a record What the heartbeat is still waiting on a PERSON for on one record, newest first, beside the vocabulary of action kinds so a caller can label one without keeping its own copy. Waiting means planned AND due — an item somebody deferred to a date is undecided but not being asked about, and stays out until that date passes. limit defaults to 50 and is capped at 200. Any record page; reading is any seat's, applying one is the admin's on its own route. GET /v1/definitions/:key/rows/:rowId/heartbeat/actions (access: def:read)

get_definitions_key_rows_rowId_intake-review

ChatGPT
Read a record's intake conflicts What an arrival (a public form today) held on this record for a person instead of writing it: a value that would have replaced one the record already holds (reason conflict, with previous and proposed), and another record the arrival also matched (reason duplicate, duplicateRowId, field names the key it matched on: ein, email, phone or legal_name). Open items (outcome review) first, then the most recently settled, at most 100. Columns this caller's policy hides and duplicates they cannot open are left out. GET /v1/definitions/:key/rows/:rowId/intake-review (access: def:read)

get_definitions_key_rows_rowId_lender-form-fills

ChatGPT
See whose blanks are open on a record's lender forms Every lender form issued on this record, newest first (at most 50): its paper (documentId), the form, its title, whether every signer has signed (executed), and with — each party who has not signed yet and still holds blanks the form leaves to them (party, open, required of those open, total), so a desk can say "3 blanks with the borrower". Never the answers themselves. Read-only; the record's own read. GET /v1/definitions/:key/rows/:rowId/lender-form-fills (access: def:read)

get_definitions_key_rows_rowId_lender-form-fills_documentId

ChatGPT
Read a form issued on a record One form issued on this record (documentId, the paper), with every value it holds field by field — source says whether the record, the sender, the signer or a later correction at the desk put it there — whether it is signed by hand (wetSignature), whether it can still be corrected (editable: every fill but one already signed electronically), and every printable copy drawn from it, newest first, each a file on the record. Read-only; the record's own read. GET /v1/definitions/:key/rows/:rowId/lender-form-fills/:documentId (access: def:read)

get_definitions_key_rows_rowId_lender-forms

ChatGPT
List a record's lender forms The newest version of every lender PDF form this record can use: those hanging on it (a lender's row holds that lender's forms, reusable on every loan sent to it) and those written against its page. Each carries its pages, its fields (each a place on a page, 0..1 from the top-left, with the record word or signing party that fills it) and how they were found (acroform, layout or none). Answers {forms, canHangForms, filledFrom}: canHangForms says another page links to this record's page (a relation column, or a finance pack link text column such as a submission's Lender), so a form uploaded here has records to fill it; filledFrom lists those pages (key, label). GET /v1/definitions/:key/rows/:rowId/lender-forms (access: def:read)

get_definitions_key_rows_rowId_lender-forms_formId

ChatGPT
Read a lender form One version of a lender form on its row, with every word it may be mapped to (the page's columns and, as <link>.<word>, each linked record's) and up to five suggested words per field. GET /v1/definitions/:key/rows/:rowId/lender-forms/:formId (access: def:read)

get_definitions_key_rows_rowId_lender-forms_formId_pages

ChatGPT
Draw a lender form's pages Each page of a lender form as a base64 PNG (drawn by processor-v2) with its size in points, for drawing fields over. A page that could not be drawn answers image: null and the note says why. GET /v1/definitions/:key/rows/:rowId/lender-forms/:formId/pages (access: def:read)

get_definitions_key_rows_rowId_lender-options

ChatGPT
Read submission lender options Reads the latest readable offer-bearing reply for this submission and the broker's primary selection. Confined to this workspace and record; private inbox replies are excluded. GET /v1/definitions/:key/rows/:rowId/lender-options (access: def:read)

get_definitions_key_rows_rowId_lender-recommendations

ChatGPT
Read this deal's lender recommendations The newest run the recommendation engine made for ONE deal, with every lender it had an opinion about — the ranked ones first with their scores and expected terms, then every lender it refused with the stage and reason it was dropped on. everRun is false when the engine has never looked at this deal, which is a different answer from a run that recommended nobody. Reach for it to answer "why was this lender not offered" about a deal in front of you; the operator's cross-tenant view of the same ledger is GET /v1/internal/lender-recommendation-runs. GET /v1/definitions/:key/rows/:rowId/lender-recommendations (access: def:read)

get_definitions_key_rows_rowId_lenders_eligible

ChatGPT
List the lenders eligible for a deal The workspace's lenders ranked for THIS deal from the platform catalogue's criteria (amount range, states served, products), each with fit eligible/partial/outside and the reasons. Use it to answer "which lenders would take this deal". GET /v1/definitions/:key/rows/:rowId/lenders/eligible (access: def:read)

get_definitions_key_rows_rowId_lifecycle

ChatGPT
Read a deal's lifecycle stage One deal's CONTRACT and FUNDING tracks: each track's current state and the history of moves that got it there. Read it before asking for a move — post_definitions_key_rows_rowId_contract and post_definitions_key_rows_rowId_funding both refuse a move that is not legal from where the track stands. A deal id from another workspace is a 404, never an empty track. Deals page only. GET /v1/definitions/:key/rows/:rowId/lifecycle (access: def:read)

get_definitions_key_rows_rowId_linked-rows

ChatGPT
Count a record's linked rows The rows of page from whose link column points at this record, counted for a rule's condition: total, statuses (live rows by their page's status value; a blank status is in total only), filled (for EVERY column of from this seat can see, how many of the rows hold a value — 0 when none do, so filled.<columnId> == total reads 'every one has it') and rows (the first 100, newest first, each {rowId, status, cells, display}). via names the link column on from (default: every one pointing at this record's page). through names a link column on THIS row to follow first, so a child row counts its siblings — a treatment row counting every treatment of its case — and record says which record was counted for (null, with zero rows, when that cell is blank or names a row this seat cannot open). Archived rows are not counted; rows behind this seat's wall on from and columns its policy hides are absent. Reach for this in a workflow read step to decide 'every linked row is X' or 'every linked row has Y filled', which the expression grammar cannot count by itself. GET /v1/definitions/:key/rows/:rowId/linked-rows (access: def:read)

get_definitions_key_rows_rowId_merchant-invitations

ChatGPT
List the invitations sent to a merchant One deal's merchant-portal invitations, newest first, LIVE AND DEAD. The dead ones are the point: that this workspace put the deal in somebody's hands, and until when, is what an audit comes looking for. The token is never here — it existed once, in the answer that minted it. Deals page only. GET /v1/definitions/:key/rows/:rowId/merchant-invitations (access: def:read)

get_definitions_key_rows_rowId_notes

ChatGPT
List a record's notes Read separate internal notes with their original author, creation time, latest editor and version. Returns newest first; pass nextCursor as cursor to load older notes. Use when reviewing notes on a lead, deal, contact or custom board record. GET /v1/definitions/:key/rows/:rowId/notes (access: def:read)

get_definitions_key_rows_rowId_offers

ChatGPT
List a deal's offers Every offer on one deal from BOTH places a lender's offer lives: offers is the deal's offer rows, side by side and best first, with the figures the desk compares them on (count and best are theirs); submissionOffers is the lender offers stated on the deal's submissions (lender, offered amount, factor, payback, term, status, the lender's response, the submission's rowId) that no offer row answers. A lender is answered once and its offer row wins. To answer 'any offers yet', name every lender in both lists - reading count alone misses an offer that sits only on a submission. A seat kept off the submissions page gets an empty submissionOffers. The POST on this same path BUILDS offers; this GET compares them. If the offers page has lost its deal relation column the call says so rather than answering 'no offers', which would be a claim about the deal. Deals page only. GET /v1/definitions/:key/rows/:rowId/offers (access: def:read)

get_definitions_key_rows_rowId_plaid_items

ChatGPT
A deal's (or merchant's) linked banks Every bank connected to this deal, with the accounts inside each one, its institution, whether it still works, and who consented to the link and when. Never the access token - there is no field for it. On the merchants page the same route lists the banks on every deal naming that merchant, each with dealId and dealName - read through the deals, never a copy. Reach for this before offering to link a bank, and to explain where a statement month with source 'plaid' came from. ONE deal's banks: to know which of MANY deals have a bank linked (or broken, or none), read the page's rows once with filter[$bank]=linked|error|none instead of calling this per deal. GET /v1/definitions/:key/rows/:rowId/plaid/items (access: def:read)

get_definitions_key_rows_rowId_plaid_monitoring

ChatGPT
Read merchant monthly monitoring Read effective monthly enrollment, the deal override, workspace default and linked bank count. Does not call Plaid. GET /v1/definitions/:key/rows/:rowId/plaid/monitoring (access: def:read)

get_definitions_key_rows_rowId_process_read-again

ChatGPT
Count what a read again would send Before post_definitions_key_rows_rowId_process with readAgain: true: how many of one record's files a read again would send (files), their total size (bytes) and how many processor jobs that is (jobs, 50 files each). A read again sends only files whose newest read failed, came back with no type, or finished before the processor's current rules (rulesSince) and was not typed by a person; never a file never read, one still being read, one in a format no reader opens, or one over 200 MB. Zero files means everything is current or being read now. GET /v1/definitions/:key/rows/:rowId/process/read-again (access: def:read)

get_definitions_key_rows_rowId_processing

ChatGPT
Read a record's document-processing status The newest processing job on one record, with its event feed in order and the documents it covered, in one query. A record nobody has processed answers a ZERO job with an empty status rather than a 404, so 'not processed' and 'bad id' stay distinguishable. What a job actually extracted is read per tab: bank-statements, tax-returns, credit-reports, or financials for the rollup. GET /v1/definitions/:key/rows/:rowId/processing (access: def:read)

get_definitions_key_rows_rowId_relationships

ChatGPT
List a record's related records One deal's cross-board graph alone — the merchant, lead, offers, submissions and other records it is linked to — for callers that want the links without the whole record. The same graph rides inside get_definitions_key_rows_rowId_detail. Deals page only. GET /v1/definitions/:key/rows/:rowId/relationships (access: def:read)

get_definitions_key_rows_rowId_script

ChatGPT
Read a record's call script The call script for ONE lead as it currently resolves: the saved script if this workspace has written one that reaches this lead, otherwise the script built from the lead's own facts. The order is row, then status, then workspace, then built. Leads page only; any other key is a 404. GET /v1/definitions/:key/rows/:rowId/script (access: def:read)

get_definitions_key_rows_rowId_signing

ChatGPT
Signing requests on a record Every signing request raised on one record, newest first - the deal's contract envelope and any other paper sent out from it. Reach for this to answer "what is out for signature on this deal, and who has not signed yet". Each entry carries per-signer state (waiting, viewed, signed, declined) and, once finished, the executed document. GET /v1/definitions/:key/rows/:rowId/signing (access: def:read)

get_definitions_key_rows_rowId_signing-events

ChatGPT
List a record's e-signature events One deal contract's envelope trail, newest first — what was sent, seen and signed, and when. Read-only here on purpose: the entries are written by the contract moves themselves, and by an external signer's webhook if one is ever wired up. There is no signing vendor in this service. Deals page only. GET /v1/definitions/:key/rows/:rowId/signing-events (access: def:read)

get_definitions_key_rows_rowId_sla

ChatGPT
Read one record's SLA state The record's chip: which stage it is in, when it entered (the newest status or stage move in its own timeline), how long it has sat in the policy's clock, the target and warning it is measured against, and when the promise comes due in wall-clock time. State is on_track, at_risk, breached, or none, and none carries a sentence saying why: the stage has no SLA, the record is settled, or the page has no status column. Reading this rings nothing. GET /v1/definitions/:key/rows/:rowId/sla (access: def:read)

get_definitions_key_rows_rowId_statements

ChatGPT
List a record's statements The bank statements filed against one record, oldest span first and unstated spans last: the files, and the spans somebody stated for them. It lists and it links, and it does NOT parse — the per-month figures (revenue, average daily balance, NSFs, positions) come from the document processor and are read with get_definitions_key_rows_rowId_bank-statements instead. GET /v1/definitions/:key/rows/:rowId/statements (access: def:read)

get_definitions_key_rows_rowId_status-history

ChatGPT
Read a record's status history One record's status ladder, newest first: each move's from, to, notes, who made it and when, derived from the record's activity feed rather than kept a second time on the row. The record's creation is the bottom rung and has no from. limit is capped at 200. Any record page. GET /v1/definitions/:key/rows/:rowId/status-history (access: def:read)

get_definitions_key_rows_rowId_submission-notes

ChatGPT
Read saved deal submission notes Returns the deal’s saved lender notes and version for authorized callers. Reuse these notes when sending to additional lenders without generating again. GET /v1/definitions/:key/rows/:rowId/submission-notes (access: def:read)

get_definitions_key_rows_rowId_submissions

ChatGPT
List the form answers filed on a record The FORM submissions filed against one record — NOT its lender submissions; a deal's sends to lenders are rows of the Submissions page (get_definitions_key_rows, params.key = "submissions", filtered by the deal). These are the answers a merchant or a person actually typed into a form, newest first, in the same shape get_forms_id_submissions answers. That route is addressed by FORM, so this is the only way to ask which submission belongs to THIS record, which is what a record's Application tab needs. Mind the noun: submissions on the POST under this row (post_definitions_key_rows_rowId_submissions_prepare-send) means LENDER submissions, a different thing entirely. A record nobody filled a form for answers an empty list; a record this workspace cannot see answers 404. limit narrows it. Answers are masked for the caller exactly as the Responses tab masks them (get_forms_id_responses, docs/forms.md § 31): an answer filling a column your column access hides, an SSN/EIN no column you can see shows, an answer to a question the form no longer asks, or every answer of an application whose record is behind your row wall is left out of answers and named in withheld. GET /v1/definitions/:key/rows/:rowId/submissions (access: def:read)

get_definitions_key_rows_rowId_submissions_contract-request

ChatGPT
Read contract request review context Returns scoped source terms, current draft, contract documents and validation issues for this submission. Reading does not approve, create a review receipt, or send. A person must open the interactive contract review to preview and explicitly send. GET /v1/definitions/:key/rows/:rowId/submissions/contract-request (access: def:read)

get_definitions_key_rows_rowId_submissions_mailer-actions

ChatGPT
List lender mailer operations Lists the latest 100 durable mailer operations for this lender submission, including delivery or reconciliation state and provider receipt IDs, without returning private payloads. Requires access to the submission record. GET /v1/definitions/:key/rows/:rowId/submissions/mailer-actions (access: def:read)

get_definitions_key_rows_rowId_subrecords

ChatGPT
List a record's subrecords Every child record hanging under this one — a deal's contacts, a merchant's locations — whichever page each is on, in section order. Each link carries the child's page key, row id, cells (masked by the CHILD page's own column policy) and the section it is drawn in. Reach for this to answer 'what does this record hold'. GET /v1/definitions/:key/rows/:rowId/subrecords (access: def:read)

get_definitions_key_rows_rowId_super-kyc

ChatGPT
Read a record's Super KYC links One record's Super KYC links, newest first, live and stopped: which sources each asked for (google, microsoft, plaid), its status (sent, opened, agreed, watching, stopped, expired), the banks and mailboxes connected, and the count of open and critical findings. Any page's record. The link's token is never here. GET /v1/definitions/:key/rows/:rowId/super-kyc (access: def:read)

get_definitions_key_rows_rowId_super-kyc_id

ChatGPT
Read a Super KYC link's findings One Super KYC link with every finding (critical first: new funder positions, agreements, statement mismatches, identity) and, when a bank is connected, twelve months of deposits and true revenue. Reach for it to answer what Super KYC found on a merchant. GET /v1/definitions/:key/rows/:rowId/super-kyc/:id (access: def:read)

get_definitions_key_rows_rowId_super-kyc_id_transactions

ChatGPT
List the bank transactions behind a Super KYC link Every bank line the link's connected banks synced, newest first, each carrying the finding raised on it when there is one. Narrow with account, direction (in or out), flagged=1 and q (payee, description or category); page with cursor. count, inMinor, outMinor and flagged are over the whole filter rather than the page, and pending lines are counted but not summed. Amounts are signed minor units: positive is money in. Reach for it to check a finding against the lines around it, or to look for something the watch did not flag. GET /v1/definitions/:key/rows/:rowId/super-kyc/:id/transactions (access: def:read)

get_definitions_key_rows_rowId_tax-returns

ChatGPT
List a record's tax returns One deal's tax returns as the processor read them: each file with the years extracted from it and the owners named in each year. A deal nobody has processed answers an empty list. Deals page only. GET /v1/definitions/:key/rows/:rowId/tax-returns (access: def:read)

get_definitions_key_rows_rowId_threads

ChatGPT
List a record's discussion threads The internal threads on one record — colleagues talking to each other about this deal or lead — most recently spoken in first, each with its roster, message count and the caller's unread count. RESOLVED threads are included: filtering them out is the caller's choice, so that 'what did we decide about this' stays answerable. Staff only; the merchant portal reaches the deal row and none of these. GET /v1/definitions/:key/rows/:rowId/threads (access: def:read)

get_definitions_key_rows_rowId_transition_requirements

ChatGPT
What a move needs first What a deal would still need before it may move to to (a status or a stage): the missing columns keyed by column id, with each one's label, kind and the sentence the move would refuse with, plus whether the target asks for a reason. Ask it before acting on a status menu, so the person fills the gap instead of reading a refusal. GET /v1/definitions/:key/rows/:rowId/transition/requirements (access: def:read)

get_definitions_key_rows_rowId_underwriting

ChatGPT
Read a deal's underwriting result One deal's underwriting record with the current version of its scorecard answers. A deal nobody has started underwriting for is a 404, NOT an empty record — the tab's 'not started' state carries a button that WRITES, so a failed read and an unstarted deal must not render the same. Start one with the POST on this same path. Deals page only. GET /v1/definitions/:key/rows/:rowId/underwriting (access: def:read)

get_definitions_key_rows_rowId_underwriting_evaluations

ChatGPT
List a deal's rule evaluations This deal's evaluations, newest first, without their per-rule results - the history strip. Each names the rule set and the version number it ran. GET /v1/definitions/:key/rows/:rowId/underwriting/evaluations (access: def:read)

get_definitions_key_rows_rowId_underwriting_evaluations_evaluationId

ChatGPT
Read one rule evaluation One evaluation in full: every rule's result, reason and observed value, the frozen version it ran, the namespace as it was, and the manual review it opened if it had one. GET /v1/definitions/:key/rows/:rowId/underwriting/evaluations/:evaluationId (access: def:read)

get_definitions_key_rows_rowId_underwriting_rule-facts

ChatGPT
List the rule namespace for a deal Every name a rule condition may read, with what it resolves to on THIS deal - average monthly revenue, NSF count, positions, tax-year figures, credit score and tradeline counts, extraction fields, and the deal's own cells. Reach for it before writing a condition, to see the numbers a threshold would be compared against. GET /v1/definitions/:key/rows/:rowId/underwriting/rule-facts (access: def:read)

get_definitions_key_rows_rowId_vendor-slots

ChatGPT
Read a record's vendor slots Which vendor handles this record's processing, legal work, credit pull or notarisation - one per kind, with the vendor's name alongside. The same seat that may open the record may read them. GET /v1/definitions/:key/rows/:rowId/vendor-slots (access: def:read)

get_definitions_key_rows_rowId_words

ChatGPT
Read a record's values as words Every column of one record in column order, each as {word, label, value}: word is the column's {{placeholder}} name (its label slugged), value the record's current value as a person reads it — an address as its line, a person as their name, a select as its label, an amount, percent or day as a paper prints it ("$100,000", "33.33%", "September 14, 2026"); "" when the record holds none. A column locked from the caller is left out entirely. What the app's right-click menu puts into a note, a task or a text typed on the record. To fill a saved message template against a record, use post_message-templates_id_preview instead. GET /v1/definitions/:key/rows/:rowId/words (access: def:read)

get_definitions_key_scripts

ChatGPT
List a page's call scripts The LEADS page's saved call scripts: the live ones first, and with ?retired=1 the whole version ladder, newest first. A script is scoped to the workspace, to a status, or to a single lead row, and that is the order they resolve in. Any workspace seat may read them — a rep has to see the words they are meant to say. Leads page only; any other key is a 404. GET /v1/definitions/:key/scripts (access: def:read)

get_definitions_key_template-words

ChatGPT
List the words a document template may use on a page Every {{word}} a document template written for this page can fill, in order, each as {word, label, group, kind}: first the page's own columns (one word per column — the pack's field name when it stamped one, else the label slugged), then the columns of every board the page links to, walked over the workspace's links (record links forward, a child board whose rows link back by one column — a case's claims — join boards, up to three hops), named <path>__<column> (claims__adjuster_name, merchant__owner_ssn) with group the board as reached ("Claims", "Loan › Borrower"). A column locked from the caller is left out; attachments are left out. A linked word fills from the OLDEST record of that board naming this one. The workspace's own words (get_message-placeholders) are not repeated here. Publishing a template refuses any word in neither list. GET /v1/definitions/:key/template-words (access: def:read)

get_definitions_key_threads_threadId

ChatGPT
Read a discussion thread One internal thread with its roster, its resolved state and the caller's unread count. Addressed by the THREAD, with no row id in the path, because a thread id already names one record and repeating it would let the two disagree. The messages are a separate read. GET /v1/definitions/:key/threads/:threadId (access: def:read)

get_definitions_key_threads_threadId_messages

ChatGPT
List a thread's messages A thread's transcript, OLDEST FIRST — the order it is read in, so no caller has to reverse a page — each message with its author's name, mentions, pinned flag and edited stamp. Deleted posts are left out. limit defaults to 100 and is capped at 500. GET /v1/definitions/:key/threads/:threadId/messages (access: def:read)

get_definitions_key_threads_threadId_participants

ChatGPT
List who is on a thread Who is on one thread, the owner first, with each person's name and when they last read it. A thread's roster is what decides who is notified of a new message. GET /v1/definitions/:key/threads/:threadId/participants (access: def:read)

get_definitions_key_transition-forms

ChatGPT
Read what this page's moves ask for Every transition form on the page: the extra fields a move ASKS for, keyed on a status value or on the lead-to-deal handoff. Separate from what a move is REFUSED without — that is the page's own gate and is not configurable. GET /v1/definitions/:key/transition-forms (access: def:admin)

get_definitions_key_workspace-access

ChatGPT
Read a pack page's workspace access The level every seat with no grant of its own gets on a pack page (Leads, Deals…): write (the default) or read. Any seat. GET /v1/definitions/:key/workspace-access (access: def:read)

get_document-requests

ChatGPT
List outstanding document requests Everything the workspace is still WAITING ON — every open document request, across every page, oldest ask first. The read for "what paperwork are we still chasing", "what stips are outstanding", "what have we not had back yet". ?limit= caps it. Only OPEN asks are here: a request leaves this list when post_document-requests_id_fulfil pins the file that arrived or post_document-requests_id_cancel withdraws it. For one record's asks read get_definitions_key_rows_rowId_document-requests instead. GET /v1/document-requests (access: workspace)

get_document-templates

ChatGPT
List document templates Every version of the paper this workspace may ISSUE — fee agreements and financing disclosures — in resolution order: the workspace's own newest first, then the system default every workspace can fall back on. ?kind= narrows to one (fee_agreement, financing_disclosure). Read it to know which version a rendered document would be made from; post_document-templates publishes the next one, and a published version's words never change. GET /v1/document-templates (access: workspace)

get_document-templates_forms

ChatGPT
List document template forms The newest version of every PDF form this workspace keeps as a document template (Settings › Document templates): each one's page whose records fill it (definitionKey, any board of any pack or a made board), its pages and fields, and how its blanks were found. Every seat reads them; only an admin uploads or changes one. GET /v1/document-templates/forms (access: workspace)

get_document-templates_forms_formId

ChatGPT
Read a document template form One version of a workspace's PDF form, with every word of its page it may be mapped to (the page's columns and, as <link>.<word>, each linked record's) and up to five suggested words per field. GET /v1/document-templates/forms/:formId (access: workspace)

get_document-templates_forms_formId_pages

ChatGPT
Draw a document template form's pages Each page of a workspace's PDF form as a base64 PNG (drawn by processor-v2) with its size in points, for drawing fields over. A page that could not be drawn answers image: null and the note says why. GET /v1/document-templates/forms/:formId/pages (access: workspace)

get_documents_id_signing

ChatGPT
Signing requests on a document Every signing request ever raised on ONE rendered document, newest first. Reach for this when you know the paper and want its signing history - the live envelope is the first entry, and a superseded one is a cancelled entry behind it. GET /v1/documents/:id/signing (access: workspace)

get_drive_files

ChatGPT
Browse the caller's Google Drive One page of the files in the Google account THIS CALLER connected in Settings - id, name, mimeType, modifiedTime, sizeBytes, webViewLink, iconLink and owner, newest first, 200 to a page with a nextPageToken to continue. ?kind= narrows to sheet, excel, csv, pdf or folder (absent means every file and no folders) and ?q= filters on the NAME; Drive's own query language is never accepted. Reach for it to find the spreadsheet a person means before importing it, or to show what is importable. Answers 409 when the caller has connected no Google account - the fix is Settings, not a retry. GET /v1/drive/files (access: workspace)

get_drive_files_fileId_export

ChatGPT
Read a Google Sheet as a file One Google Sheet from the caller's own connected account, rendered as ?format=xlsx (base64, the whole workbook) or ?format=csv (plain text). ?gid= names one TAB, which only csv can honour - an xlsx export already carries every tab. Reach for it once a sheet has been chosen and its rows are what you need; the import routes take it from there. Refuses over 64 MB rather than truncating, because a cut spreadsheet still imports and silently loses rows. GET /v1/drive/files/:fileId/export (access: workspace)

get_drive_files_fileId_overview

ChatGPT
Glance at a Google Sheet's tabs One Google Sheet from the caller's own connected account at a glance, without exporting it: for each of its first ten grid tabs, in workbook order, the tab's name, its grid size (rowCount and columnCount as Sheets states them, empty rows included), its first row as header and up to five rows under it as sample, every cell as the sheet shows it. Reach for it to show which tab holds what, or to check a sheet's columns before importing; get_drive_files_fileId_export reads the rows themselves. Answers 409 when the caller has connected no Google account, and Google's own 4xx when the file is not a Google Sheet. GET /v1/drive/files/:fileId/overview (access: workspace)

get_email-blasts_id

ChatGPT
Read one email blast One EMAIL blast by id — the blast itself, its audience and outcome counts, and a sample of up to 25 recipients. The kind comes from the PATH, so an SMS blast's id is a not-found here and get_sms-blasts_id is its twin. This is the cheap read for an opened blast; reach for get_blasts_id_report instead when the question needs the whole audience, the lines the copies went out on, and who replied. GET /v1/email-blasts/:id (access: workspace)

get_email_blast

ChatGPT
For non-merchant Specter broker/operator users only. Reads one email blast in full: what was sent, to how many, and how each recipient landed. Read-only.

get_facility_business_research

ChatGPT
Gets the stored business research for one application/facility: owner name, registration state, initial filing date, Secretary of State active status and good standing, court records and legal actions (business and owner), research notes, source URLs, and AI-synthesized risk signals. Use for '/research', 'what do we know about this business', 'is this business legitimate', 'any liens or court records'. Reads stored research only — it does not run a new search. It is read-only.

get_facility_credit_reports

ChatGPT
Lists the stored credit reports for one application/facility. Use when the user asks about credit, scores, or the owner's credit profile on a deal. Reads stored pulls only — it does not order a new credit report. It is read-only.

get_files

ChatGPT
List the workspace's files Every file the caller may open, newest first, at most 2,000: the workspace-loose uploads AND the files filed on records the caller can open — a file on a record behind their wall is not listed. Each carries links (the record it is filed on as id "pin", then every record it is linked to, each with definitionKey, rowId and the record's label; records the caller cannot open are left out) and starred, the caller's own star. Metadata only: name, content type, size, who uploaded it or which address it arrived from, and how it is classified. Soft-deleted files are never listed. GET /v1/files (access: workspace)

get_files_id_download

ChatGPT
Download a file Mint a short-lived signed download URL (15 minutes) for one file by id, loose or pinned to a record, and answer it with the file's name and content type. The bytes and the storage path never leave the server. Refused with 409 while a document is locked waiting for its unlocked canonical PDF to be saved. GET /v1/files/:id/download (access: workspace)

get_files_id_reading

ChatGPT
Read what was read off one file Everything the processor made of ONE file, from the newest job that read it (whichever job that was): documentType and documentTypeSetBy, landed, attempt (its newest processing attempt in any job, null when never sent), readByJobId, extractions — every field filed, verbatim (field, value as the engine reported it: text, number, list or object, page, confidenceBps), grouped by kind, with the engine's notes (the document's text) — and overview: summary, keyFacts [{label, field, value, page, sensitive}], parties, dates, amounts, proposedType. sensitive lists the paths holding a masked whole identifier ("••••1234"); the whole value is never in this read. Works for a loose file and for a file read in an earlier run. A file whose fields are empty was not read, or the processor stated nothing. Unknown, deleted and walled files are a not-found. For every file of a record at once, get_definitions_key_rows_rowId_document-facts. GET /v1/files/:id/reading (access: workspace)

get_files_id_result-changes

ChatGPT
Read what a document's results did to the record What the board's result mappings did with this document's latest processed result: summary {applied, review, kept} and each change — the field, the record and column it was for, the value read (extracted), the value for the column (proposed), the value it replaced or would replace (previous), and outcome applied, review (held for a person: reason rule or conflict), kept (left on the document: filled, same, no_link, unreadable, column_missing, column_changed, superseded), accepted or rejected. Reviews still open from an earlier run are included. Changes on records the caller cannot read are left out. GET /v1/files/:id/result-changes (access: workspace)

get_files_id_shares

ChatGPT
List a file's live share links The share links of one file that still open — not revoked, not ended by a delete or a replace, not past their week — newest first, each {id, fileId, createdBy, createdAt, expiresAt}. createdBy is the person id of whoever made the link; anyone who can open the file sees every live link on it, not only their own. The link's address is never in this answer: it exists once, in the answer that minted it (post_files_id_share). Revoke one with delete_files_id_shares_shareId and its id. An unknown, deleted or walled file is a not-found. GET /v1/files/:id/shares (access: workspace)

get_finix_transfers

ChatGPT
Read the money moved through Finix Every debit and credit this workspace has attempted, newest first, narrowed by deal, payment or payout. Reach for it to answer "did that collection actually go through" or "where has that withdrawal got to": each row carries the rail's own state, its trace number and, on a failure, Finix's reason. Filtering by deal also returns that deal's merchant, so one call says both whether the deal is on the rail and what it has moved. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. GET /v1/finix/transfers (access: workspace:books)

get_forms

ChatGPT
The workspace's forms Every form this workspace has built — drafts as well as published — each with its fields, sections, settings, slug and status. Read this to find a form's id before reading its answers, its link or its analytics. GET /v1/forms (access: workspace)

get_forms_id

ChatGPT
Read one form One form whole: its questions with the record column each maps to, its sections, its settings, its slug and whether it is published. Read it before patching — a PATCH replaces fields, sections and settings entire, so you have to hold the current list to change one line of it. GET /v1/forms/:id (access: workspace)

get_forms_id_analytics

ChatGPT
Form analytics What one form actually did over a window (7d, 30d or 90d): views and distinct visitors, drafts started, applications completed and corrections, how many were signed, the median time to complete, a per-question drop-off table, conversion split by the ?src= label on the link, and a daily series. Reach for it to answer 'is this form working', 'where do people give up' or 'which channel converts'; use GET /v1/forms/:id/submissions for the answers themselves. Two things the numbers do NOT mean: dropOff names the last question somebody TOUCHED before abandoning a saved draft, which is not the question that put them off and not one they read and skipped, so do not report a cause from it; and 'starts' counts saved drafts, so a form whose saveDrafts setting is off shows few or none — that is a setting, not a finding. The VISIT figures, over the visits that began in the window: completionRate (visits that filed / visits), medianFillMs (median duration of a visit that filed, on the page's clock; null when none was measured), devices and visitReferrers (sessions and how many filed, by device class and by referring site — referrers is the older views-and-filings table by referrer), and fieldFunnel (per question in the form's order: visits that reached it, completed it, dropped, and the drop rate). Who those visitors were is get_forms_id_visits. GET /v1/forms/:id/analytics (access: workspace)

get_forms_id_responses

ChatGPT
The applications one form took ONE form's applications, newest first, one line each: what stands now (the newest correction's answers), the records it filed into (the deal and its merchant, with names as you may read them), the rep its link credits, whether it was signed, how many times it was corrected, its source and UTM words, its application PDFs and how many edit links stand open. Answers a column you may not see fills are withheld and named in withheld. ?q= searches the text, email and phone answers (a business name, an email address); ?limit= (default 50, at most 200) and ?offset= page it, and total counts every match. The trail of every row, corrections included, is get_forms_id_submissions. GET /v1/forms/:id/responses (access: workspace)

get_forms_id_responses_submissionId_visit

ChatGPT
The visit that filed one application How ONE application's visitor got there and what they did before filing: the visit (first and last seen, duration on the page's clock, how far they got by question and page, whether they reached submit, the outcome, the address, user agent, referrer, language, device, browser and OS the request showed, validation errors hit, uploads, and the questions that refused them) and its timeline of events (question focused, completed, cleared, page advanced, upload, submit attempted, submitted), each naming the question by id and label — never an answer. Address it by the form id and the application id from get_forms_id_responses. session is null for an application filed before visits were kept or while the form's analytics were off; an application this form does not hold is a 404. Every visitor, filed or not, is get_forms_id_visits. GET /v1/forms/:id/responses/:submissionId/visit (access: workspace)

get_forms_id_submissions

ChatGPT
What people filed through one form What people filled in on ONE form — the answers filed through it, newest first, ?limit= to cap it — what a merchant or a visitor actually typed. This is the FORM's side of the question; the RECORD's side is get_definitions_key_rows_rowId_submissions, the same shape addressed by the record the answers landed on. Answers are masked for the caller exactly as the Responses tab masks them (get_forms_id_responses, docs/forms.md § 31): an answer filling a column your column access hides, an SSN/EIN no column you can see shows, an answer to a question the form no longer asks, or every answer of an application whose record is behind your row wall is left out of answers and named in withheld. GET /v1/forms/:id/submissions (access: workspace)

get_forms_id_visits

ChatGPT
Everyone who visited one form The People list for ONE form: every visit that BEGAN in the window, newest activity first, filed or not — when, for how long, how far they got, the outcome, and the address, device, browser, OS, language and referring site the request showed, plus submitterLabel (the record the application filed into, named as you may read it) for a visit that filed. ?outcome= narrows to submitted, failed_at_submit (pressed submit and never filed), abandoned (touched a question and left) or opened_only; ?window= is 7d, 30d or 90d (default 7d), the same windows as get_forms_id_analytics; ?limit= (default 50, at most 200) and ?offset= page it, and total counts every match. One application's visit with its event timeline is get_forms_id_responses_submissionId_visit. GET /v1/forms/:id/visits (access: workspace)

get_fos_setup_catalog

ChatGPT
Tenant-admin-only setup helper. Lists which manual settings are supported for Deals, Leads, and Merchants, including group and status setup endpoints. Use in the admin sidepanel before creating setup records. Do not use inside Workflow Builder automation design.

get_full_workup

ChatGPT
ONE-CALL deep briefing for a single facility/application: everything in get_underwriting_packet (enriched deal, submissions, offers, bank statements, timeline) PLUS statement periods, tax years, the offer comparison, and a communications snapshot (latest channel + recent messages). PREFER THIS over get_underwriting_packet followed by separate period/tax/offer/comms reads when preparing a full underwrite or /analyze. Read-only. Sections may carry a section_errors entry instead of failing the whole workup.

get_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Reads ONE Heartbeat action in full — its payload (for a meeting: title, start, end, and the complete guest list with each guest's email), its status, and its acceptance preview. Use this to answer questions about a specific proposed action, such as who is invited to a meeting or what a field change would set, before or after accepting it. This operation only reads one stored proposal record; it never executes its payload or registers additional tools.

get_integrations

ChatGPT
Read integrations Every third-party wiring this workspace has — provider, status, the public settings bag, and for each credential only whether one is stored and its tail — plus the closed list of providers so a caller can offer one that has no row yet. ADMIN ONLY, reading included: an integration row is the workspace's credentials. No credential value is ever in this answer. GET /v1/integrations (access: workspace:admin)

get_integrations_managed-zoom_draft

ChatGPT
Read managed phone setup Read the workspace admin's saved preparation for a Specter-managed Zoom line. Revision 0 means no draft. Help the user explain their actual business, messaging purpose, sample message and opt-in process. This is preparation only; never claim that a draft is a submitted or approved 10DLC brand/campaign, purchased number or active service. GET /v1/integrations/managed-zoom/draft (access: workspace:admin)

get_integrations_managed-zoom_hosted

ChatGPT
Read hosted 10DLC onboarding pages Read this workspace's hosted SMS consent, privacy and terms package and current publication states. Null means no package. Stale means the draft or page content changed; regenerate and review before publishing. Page publication is not website-link verification, carrier submission, campaign approval or permission to send. Read the managed Zoom review separately for dated provider evidence. GET /v1/integrations/managed-zoom/hosted (access: workspace:admin)

get_integrations_managed-zoom_offers

ChatGPT
Read verified managed phone prices Workspace admins only. Read current unexpired monthly retail unit prices for the configured managed Zoom account. An empty list means no verified current pricing. Revision IDs identify immutable price revisions. Taxes and usage are additional; 10DLC assistance is free. This does not reserve a price, order a line, start checkout or activate service. Never substitute estimates when a verified price is absent. GET /v1/integrations/managed-zoom/offers (access: workspace:admin)

get_integrations_managed-zoom_orders

ChatGPT
Recover managed phone orders Workspace admins only. Read up to 50 saved orders in this workspace, prioritizing unfinished orders and then newest creation time. Use when reopening setup or returning from payment checkout. Available even when new purchases are disabled. This read does not contact providers, charge, provision, or activate service; use the individual order status for billing evidence. GET /v1/integrations/managed-zoom/orders (access: workspace:admin)

get_integrations_managed-zoom_orders_id

ChatGPT
Read managed phone order status Workspace admins only. Read a saved managed phone order and its recorded billing evidence. Use after checkout or when explaining payment and subscription status. Works even when new purchases are disabled. Performs no provider calls or purchases. Payment recorded and provider active status do not establish phone activation; retain deletionSeen when interpreting later observations. Unknown or foreign orders are not returned. GET /v1/integrations/managed-zoom/orders/:id (access: workspace:admin)

get_integrations_managed-zoom_review

ChatGPT
Read managed phone review Read the latest internal Specter review and its feedback. A prepared request is not a carrier approval or an active phone line. Workspace admin only. GET /v1/integrations/managed-zoom/review (access: workspace:admin)

get_integrations_plaid_partner

ChatGPT
Read the Plaid onboarding The workspace as an END CUSTOMER of Specter's Plaid partner account, as last heard from Plaid: its client id, Plaid's status, the two OAuth reviews, each bank's registration and the answers it applied with. configured is false and customer null when it never onboarded. Never a secret — only whether Plaid has issued one. Calls nobody; refresh is the route that asks Plaid. Admin only. GET /v1/integrations/plaid/partner (access: workspace:admin)

get_integrations_plaid_readiness

ChatGPT
Read Plaid readiness Whether a bank can be linked on this workspace right now, on whose Plaid keys (its end customer's under Specter's partner account, the keys it typed, or Specter's own for merchant-facing links), and on what terms: free during the trial, counted against the bank login package (included logins used, overage so far, when the period ends), blocked once the trial has ended without the package, or not counted at all on the workspace's own Plaid account. Calls nobody. Admin only. GET /v1/integrations/plaid/readiness (access: workspace:admin)

get_integrations_provider

ChatGPT
Read integrations provider One provider's wiring: its public config, its status, and which credentials are on file (stored-or-not plus a tail, never the value). Admin only. A provider with no row is a not-found, which is how a caller tells 'never configured' from 'configured and paused'. GET /v1/integrations/:provider (access: workspace:admin)

get_integrations_ringcentral_status

ChatGPT
Read RingCentral connection status Read the caller's own RingCentral connection, last sync, recording retry errors and workspace-AI choice. Answers configured plus a nullable grant; null means no connection. History imports do not require a Finance Pack. GET /v1/integrations/ringcentral/status (access: workspace)

get_integrations_ringcentral_users

ChatGPT
Read RingCentral account users and lines Workspace administrators can read users and assigned phone numbers from their own connected RingCentral account. Active workspace members are matched by email for display only; this does not change communication access or ownership. GET /v1/integrations/ringcentral/users (access: workspace:admin)

get_integrations_zoom_phone-numbers

ChatGPT
Side effects: Refreshes the caller’s stored Zoom line inventory, updating line properties and removing lines no longer assigned by Zoom. Read integrations zoom phone-numbers The caller's own Zoom Phone lines, refreshed from Zoom through their own grant and stored. When Zoom cannot be reached it answers the lines we last saw with stale=true and a sentence saying so. Not connected answers an empty list. Any seat. GET /v1/integrations/zoom/phone-numbers (access: workspace)

get_integrations_zoom_status

ChatGPT
Check whether Zoom is connected Whether YOUR Zoom is connected, with your grant's details and the line your texts go out from. Never a colleague's connection. Not connected is an ordinary answer rather than an error, so this is what a pane asks before offering to connect. Any seat. GET /v1/integrations/zoom/status (access: workspace)

get_integrations_zoom_users

ChatGPT
Read integrations zoom users The Zoom Phone users on the connected account, live, each with the lines Zoom assigned them and the workspace member whose email matches. Derived on every read; nothing is stored. GET /v1/integrations/zoom/users (access: workspace:admin)

get_iso-network_disclosures

ChatGPT
Read the ISO network roster The brokerages Specter has disclosed to this lender: legal and trading name, parent brokerage, relationship and verification status, onboarding date, the basis for the disclosure, and when it was made and announced. Nothing else — no tax id, owner, application or other lender's relationship. enabled: false with no items means this workspace is not on the network, which is a different answer from an empty roster. ADMIN OR OWNER; every read of a switched-on roster leaves an activity line. GET /v1/iso-network/disclosures (access: workspace:admin)

get_lead

ChatGPT
For non-merchant Specter broker/operator users only. Returns one lead's full record — contact details, business information, status and tags. The free-text notes field is NOT shown to you, and no tool reads a lead's notes back — never claim to have read them or to know what they say. Use this after search_records has resolved a name to an id, and ALWAYS before update_lead or set_lead_status: those change a live record, and this is how you see the current values so you do not overwrite something a person entered. It is read-only.

get_leads_snapshot

ChatGPT
Lead snapshot New, contacted, qualified and converted lead counts over a window, plus the board total. Each is a dated fact and not a status word: new is the created cell, contacted the last-activity cell, qualified the submitted cell, and converted means a deal was opened from that lead inside the window. A column the workspace removed is named in unanswerable instead of counting zero. Use it for the top-of-funnel picture. GET /v1/leads/snapshot (access: workspace)

get_lender-directory

ChatGPT
List every lender and how a submission reaches it The workspace's own Lenders board as destinations: every entry with how a submission would reach it (a share to that lender's Specter workspace, or the package by email to its address) and whether it is reachable at all. Use it before submitting a deal to see who can be sent to. GET /v1/lender-directory (access: workspace)

get_lender-workspaces

ChatGPT
List the lenders linked to a Specter workspace Which entries on our Lenders board are linked to a Specter workspace — the lenders a deal can be submitted to IN-NETWORK rather than as e-mailed paper. Only the linked ones, which is the small half of the list; get_lender-directory answers every lender on the board with how a submission would reach it. put_lender-workspaces_rowId makes or corrects a link. GET /v1/lender-workspaces (access: workspace)

get_lender_submission

ChatGPT
Gets one lender submission visible to the authenticated user. Use when the user asks for lender submission details, current status, requested terms, offered terms, approved terms, or lender response context. It is read-only and tenant/user scope is enforced by the backend.

get_lenders_id_analytics

ChatGPT
Lender analytics The submission analytics narrowed to ONE entry on this workspace's Lenders board: how many crossings were sent to that lender in the window, its approval and decline rates, how long it takes to answer, and its average offer. The id is the lender's row id on the Lenders board; another workspace's id is a 404. Use it to compare one lender against the workspace's overall figures. Never once per lender to rank them: get_submissions_analytics answers every lender in one call, and with source=page ranks them over the Submissions page's rows. Many at once: get_submissions_analytics GET /v1/lenders/:id/analytics (access: workspace)

get_mail-connections_id_messages_messageId_appearances

ChatGPT
Find a message in mailbox folders Read or choose native Yahoo/AOL folders for your own connected workspace mailbox. Requires the current mailbox owner and active consent. Folder names and derived data are prohibited from advertising use. GET /v1/mail-connections/:id/messages/:messageId/appearances (access: workspace)

get_mailboxes

ChatGPT
Read the workspace's connected mailboxes Every mailbox this workspace has connected: the provider (gmail or outlook), the address, whether the connection is proven, and when it last synced. A MAILBOX HERE IS A WATCHED INBOX — the account mail arrives at — not a folder of mail and not a thread: to read what was said, get_comms_conversations and get_comms_texts. It is not the list of addresses anybody may send from: a seat's own connected email account sends without being listed here, so an empty list does not mean email cannot go. "What can I send from" is get_comms_senders. GET /v1/mailboxes (access: workspace)

get_me_attention

ChatGPT
Read what needs my attention A live, bounded work queue for the signed-in person: unresolved missed calls and unanswered messages in readable conversations, their overdue tasks, and material updates from the last seven days on assigned records. Reading a thread does not resolve it. Failed or queued sends do not count as replies. Each item carries its exact conversation, task or record target. Up to 20 per category; hasMore means the selection or the 200-event update scan was capped, not an exhaustive total. No message bodies or changed field values are returned. GET /v1/me/attention (access: workspace)

get_me_calendar

ChatGPT
Read your calendar settings The caller's own calendar preferences: which calendar new meetings land on, their working hours (minutes from local midnight), the default meeting length, whether a task with a due TIME is mirrored onto the calendar, what happens to that meeting when the task is completed, and how long before a meeting the bell rings. Answers the defaults rather than 404 when nothing has been saved; \stored\ says which. GET /v1/me/calendar (access: workspace)

get_me_context

ChatGPT
Side effects: The seat summary includes calendar counts, reconciling canceled meetings and mirrored tasks and sending due reminders. Read everything about who is asking Who is asking, in one read: the person (name, email, phone, role, timezone, workspace), the lines they may send from — their own and those shared with them, default on the text line a send naming no line uses — their connected mailboxes and calendars, the pages and boards they can open with row counts, the workspace's teams and their leads, the areas the workspace set AI (Neo and connected AI apps) to Never, and today's counts. Reach for it before asking which line, mailbox or board they mean. Pass timeZone (IANA) to cut today at their midnight. A part that could not be read is named in problems; the rest still answers. GET /v1/me/context (access: workspace)

get_me_day

ChatGPT
Side effects: The day summary reads connected calendars, reconciles canceled meetings and mirrored tasks, and sends due reminders. Read my day THE FIRST CALL for "what does my day look like", "what's on today", "what do I need to do today", "anything I'm missing" — one read answering all three halves of it: the caller's own meetings today, the tasks on their plate that are due today or already overdue, and what changed since yesterday on the records they are assigned to. Prefer it over calling get_calendar_events, get_tasks and get_activity separately: those three answer one piece each and each one costs a round trip (the same question took five tool calls and 17 seconds before this route existed). Pass timeZone as an IANA name (Europe/London) so the day is cut at the person's own midnight; with none it answers a UTC day and says so. Reach for the three separate reads instead when the question names a window other than today, somebody else's work, or one particular record. GET /v1/me/day (access: workspace)

get_me_getting-started

ChatGPT
Read my Getting Started checklist The signed-in person's Getting Started card: one row per setup step, each read from what actually happened, never ticked by hand. Rows, in order: account (always done), email (a mailbox they connected syncs here), records (the workspace holds a deal, merchant or lead), texting (their Messages reached the app; only when device is mac), team (somebody else holds a seat; only when the wizard's role was brokerage), submission (they sent a deal to a lender). Each row says done and when it first became true. completedAt is when they finished the card, null while it is still theirs to do. device is required: mac, windows or web. GET /v1/me/getting-started (access: workspace)

get_me_recent-records

ChatGPT
Read the records I opened recently THE FIRST CALL for "what was I just looking at", "that deal I had open", "the lead from earlier": the signed-in person's own latest distinct records, newest first — eight by default, limit up to 25. Each comes back in a search hit's terms: its surface and key (a board's is board-<hex>, beside its boardId), the page or board's name, the rowId, its label, where it stands (status, statusLabel) and when it was opened (openedAt). Every record is re-read under the person's access as it stands now, so one they can no longer open, or that was archived or deleted, is simply not in the answer. One person's own history, read off the session; nobody else's opens are ever answered. GET /v1/me/recent-records (access: workspace)

get_me_screen-capture

ChatGPT
Read my screen-capture protection Whether the signed-in person's desktop window is protected from screenshots and screen recordings, how, and until when it is not: mode is on (protected, and they may grant themselves a window), locked (protected, and they may not) or off (not protected); style is hidden (the windows are excluded, so a screenshot and a recording both come back empty) or blur (the app frosts itself while a recorder is running, which covers a recording but NOT a screenshot, because macOS never tells an app one was taken); allowedUntil is set while a self-granted window is still running. Never 404s — somebody nobody has decided about reads on and hidden, the protected default. One person's own state, read off the session. GET /v1/me/screen-capture (access: workspace)

get_me_script-template

ChatGPT
Read my call-script template The signed-in person's own cold-call script: the spine they read top to bottom, the objection tree under it, and the tone the generator writes in. Never 404s — someone who has never saved reads the default, so this always answers a script. One person's row; it says nothing about anyone else's. GET /v1/me/script-template (access: workspace)

get_me_usage_summary

ChatGPT
Read my usage summary The signed-in person's own footprint over a 7d, 30d or 90d window: what they touched, per feature, per day, with their active minutes. One person's figures — the actor comes from the session, so this never answers for a colleague. The whole workspace is get.workspace.usage.summary, which is admin only. GET /v1/me/usage/summary (access: workspace)

get_message-automation-rules

ChatGPT
Read the standing message rules The standing rules this workspace has armed — each one "whenever this happens to a record of this kind, write from this template" — with the trigger event, the record kind it watches, the template it names, whether what it produces needs approving, which field carries the address, and whether it is switched on. ?channel=email|sms. A rule is a standing instruction, never a message: nothing here says what was sent (get_comms_texts) or what is waiting to be approved (get_message-drafts). GET /v1/message-automation-rules (access: workspace)

get_message-drafts

ChatGPT
Read the approval queue Read saved messages. With inbox=true, returns only the caller's own unsent email drafts, excluding Heartbeat and queued emails, for Inbox → Drafts. Supports limit (50 default, 200 maximum), offset, status and channel. Without inbox=true this is the workspace approval queue; user-created emails remain visible only to their author. An approved draft is not necessarily sent: detail.queuedMessageId and detail.queuedConversationId are the durable queue receipt when its send door was used; delivery is read from the conversation. GET /v1/message-drafts (access: workspace)

get_message-placeholders

ChatGPT
List the placeholders a message template can use The placeholder vocabulary a message or document template may draw on — the tokens that resolve against a record when a template is rendered, and what each one stands for. Read it before writing a template, so a token that resolves to nothing is never sent to a merchant. Read-only for everyone: growing this list is a schema change, not a write. GET /v1/message-placeholders (access: workspace)

get_message-templates

ChatGPT
Read the message templates The shelf of reusable message bodies this workspace keeps, email and text in one list: name, category, the four parts (subject, greeting, body, closing) with their {{placeholders}} still in them, which one is the default for its category, and whether it is still offered in the picker. ?channel=email|sms, ?category=, ?activeOnly=true, ?limit=. A template is words with holes in them, nothing that has gone anywhere — post_message-templates_id_preview fills one in against a record to show what it would say, post_message-drafts queues a real letter for somebody to approve, and post_blasts aims the same words at a whole list of people. GET /v1/message-templates (access: workspace)

get_my_calendar_settings

ChatGPT
For non-merchant Specter broker/operator users only. Which calendar the signed-in user's meetings and follow-ups land on: their preference (auto | google | outlook), what it resolves to right now, and whether Google Calendar and their Outlook mailbox are connected — outlook.needs_reconnect means the mailbox was connected before calendar access was part of the grant. Use it to explain why a calendar read or booking was refused, and to tell the user which setting to change (Settings → Calendar; Outlook connects under Settings → Integrations → Outlook). Read-only; the preferences are changed by the user on the page, not through this tool.

get_my_daily_brief

ChatGPT
For non-merchant Specter broker/operator users only. The signed-in user's DAY in one read: the meetings still ahead, tasks due today, anything overdue, leads whose follow-up date is today, how many Heartbeat actions are waiting, and the open windows left to book into — with a one-sentence summary. This is what /digest should answer from: one call, instead of stitching together list_tasks, list_leads and list_heartbeat_actions. Pass day (today / tomorrow / a weekday name / YYYY-MM-DD, in the user's own time zone) to brief a different day; default today. When calendar_connected is false, repeat calendar_reason to the user verbatim — the rest of the brief is still real. Read-only.

get_my_specter_profile

ChatGPT
Reads the authenticated user's Specter profile, active tenant, and role context. Use before user-specific application, facility, or submission requests to confirm identity and authorization. Requires a connected Specter account or bearer token; it does not mutate data.

get_neo_memories

ChatGPT
Read what Neo remembers about you What Neo remembers about the CALLER across conversations — the answer to "what do you remember about me", and the list to read for an id before changing or forgetting one. Rules (how they want Neo to work) come first, then facts (about them or their desk), newest first within each; each carries id, kind, text, createdAt, updatedAt and sourceThread, the conversation it was saved in (null when typed in Settings or brought over from their old note). Only the caller's own are ever listed. Every conversation already starts with them, so read this when asked, not to recall them. GET /v1/neo/memories (access: workspace)

get_notification-preferences

ChatGPT
Read your notification settings The caller's own choice for each notification type — whether it shows in the app and whether it makes a sound. Read this before changing one. GET /v1/notification-preferences (access: workspace)

get_notification-types

ChatGPT
Read the notification types Every kind of notification this workspace can raise, Specter's own and the workspace's, each with its label, category, priority and defaults. The list the settings screen draws a row per, and the keys a raise names. GET /v1/notification-types (access: workspace)

get_notifications

ChatGPT
Side effects: Sweeps due tasks and creates missing task-due notifications before listing notifications. List notifications The person's notifications, paged, with the unread total. GET /v1/notifications (access: workspace)

get_notifications_unread-count

ChatGPT
Side effects: Sweeps due tasks and creates missing task-due notifications before counting unread notifications. Count your unread notifications How many unread notifications the caller has, for the bell's badge. Also sweeps tasks that have come due, so an overdue task is noticed by this read rather than by a scheduler. GET /v1/notifications/unread-count (access: workspace)

get_obligations

ChatGPT
What is owed on the deals The things somebody still has to do or hand over on a deal — a repayment, a document to submit, a stipulation, reporting, a covenant check, insurance, a licence renewal, a review — soonest due first, each with who is responsible, who it is assigned to, its priority, its amount and its SLA clock. Narrow with ?facilityId=, ?dealRowId= (a deal record on our own board), ?assignedToUserId=, ?status= (pending, due, completed, overdue, waived, cancelled), ?priority= (low, medium, high, critical) and ?limit= (100 by default, 500 max). For the rows as a PAGE — the whole Obligations board with its columns, saved views and facet filters, or one deal's rows through ?dealRowId= — read get_definitions_key_rows with key obligations; this route is the servicing operation itself. GET /v1/obligations (access: workspace)

get_obligations_id

ChatGPT
Read one obligation One obligation in full: what is owed and by whom, its type, priority and status, the deal it hangs off, who it is assigned to, the amount in minor units, when it is due, the document type it is waiting on, its SLA clock and breach severity, and the notes against it. An obligation belonging to another workspace answers not found. GET /v1/obligations/:id (access: workspace)

get_offer

ChatGPT
For non-merchant Specter broker/operator users only. Gets one offer by ID: full terms, fees, commission, counter values, conditions, and the presented/accepted/declined/expired timestamps. Use after list_offers or compare_application_offers has identified which offer matters. Read-only.

get_onedrive_files

ChatGPT
Browse the caller's OneDrive One page of the files in the Microsoft account THIS CALLER connected in Settings - id, name, mimeType, modifiedTime, sizeBytes, webUrl and owner. With no ?q= it is their recent files; ?q= searches the whole drive by name (quoted here, never OData). ?drive= (a library id from get_onedrive_libraries) lists that SharePoint library instead - its top folder with no ?q=, a search of every folder with one - and its file ids carry the library, ready for the content route. Folders are left out. ?pageToken= continues with the nextPageToken the last page gave, with the same ?drive=. Reach for it to find the Excel workbook, PDF or csv a person means before attaching it. Answers 409 when the caller has connected no Microsoft account, or (with ?drive=) one that cannot read SharePoint - the fix is Settings, not a retry. GET /v1/onedrive/files (access: workspace)

get_onedrive_libraries

ChatGPT
List the caller's SharePoint libraries The SharePoint document libraries of every site THIS CALLER follows, through the Microsoft account they connected in Settings - id, name, site and webUrl. Each id is what get_onedrive_files takes as ?drive= to browse that library. ?pageToken= continues with the nextPageToken the last page gave (pages are sites). Reach for it when the file a person means lives on a team site rather than their own OneDrive. Answers 409 when the caller's Microsoft connection cannot read SharePoint sites - they reconnect it in Settings, not retry. GET /v1/onedrive/libraries (access: workspace)

get_onedrive_shared-workbook

ChatGPT
Read the workbook behind an Excel share link The workbook a OneDrive or SharePoint share link points at (?link= - onedrive.live.com, 1drv.ms or any *.sharepoint.com address), read through the Microsoft account THIS CALLER connected in Settings. Answers {name, contentType, contentB64}. Any other host is refused with 400 before Microsoft is asked; a link to a document that is not .xlsx, .xlsm, .xlsb, .xls or .csv is 422; a link the caller's account cannot open is 404 saying it has to be shared with them; over 20 MB is 413. Answers 409 when no Microsoft account is connected - the fix is Settings, not a retry. Reach for it when a person pastes an Excel Online link to import. GET /v1/onedrive/shared-workbook (access: workspace)

get_payment-schedules

ChatGPT
The repayment schedules being collected The schedules this workspace is collecting on — what was funded, the total payback, the instalment amount and cadence, how much has been collected, what is still outstanding and whether the schedule is behind — next payment date first. Narrow with ?facilityId=, ?dealRowId= (a deal record on our own board), ?status= (active, paused, completed, defaulted, settled, cancelled), ?delinquent=true|false and ?limit= (50 by default, 200 max); the CRM's separate active and delinquent lists are this one filtered. For the rows as a PAGE — the whole Payment schedules board with its columns, saved views and facet filters, or one deal's rows through ?dealRowId= — read get_definitions_key_rows with key payment-schedules; this route is the servicing operation itself. GET /v1/payment-schedules (access: workspace)

get_payment-schedules_id

ChatGPT
Read one repayment schedule One schedule in full: funded amount, total payback, factor and interest rate, fees, the instalment amount and frequency, start and expected end, how much has been collected, what is outstanding, how many payments remain, the next payment date, and whether it is delinquent. All money in exact minor units, in fields named …Minor. Reach for get_payment_schedules_id_reconciliation when the question is how the schedule is performing rather than what its terms are, and get_payment_schedules_id_payments for the instalments themselves. A schedule belonging to another workspace answers not found. GET /v1/payment-schedules/:id (access: workspace)

get_payment-schedules_id_payments

ChatGPT
The instalments under one schedule Every instalment on one schedule, in payment-number order: what is due, when, what actually landed, the variance, any penalty, and the state each one is in (scheduled, pending, processing, completed, partial, failed, missed, waived, reversed). ?limit= caps the page (500 by default, 1000 max). This is the schedule's own run; for what is due across the WHOLE book reach for get_payments_overdue or get_payments_upcoming, and for the Payments board as a page read get_definitions_key_rows with key payments. GET /v1/payment-schedules/:id/payments (access: workspace)

get_payment-schedules_id_reconciliation

ChatGPT
How one schedule is performing The schedule's report, computed at read time from the instalments themselves: total expected, total received, what has been missed, what has been waived, penalties, the variance, the collection rate, how many payments were made, missed, waived and still unreconciled, plus the money split by allocation type (principal, interest, fee, penalty, commission, buyout, other). Nothing is stored, so nothing can be stale. Reach for it to answer how much of this deal has actually been collected and what is still outstanding; get_payment_schedules_id has the TERMS, and get_portfolio_aging asks the same question across the whole book. GET /v1/payment-schedules/:id/reconciliation (access: workspace)

get_payment-schedules_id_servicing-authorizations

ChatGPT
Side effects: Expires lapsed servicing authorizations when the caller has ledger write access. A schedule's debit authorisations The merchant's signed permissions to debit this schedule — the legal evidence that taking the money was allowed — newest first, with the channel and address the request went to, the amount and frequency authorised, the exact words the customer was shown, the three consents, when and from what IP it was accepted, and why it was closed. ?limit= caps the page. Anything pending whose window has closed is retired to expired before the answer is built, so a pending row here is genuinely still pending. These are PERMISSION records, not money: for the instalments and what has been collected reach for get_payment_schedules_id_payments. GET /v1/payment-schedules/:id/servicing-authorizations (access: workspace)

get_payments_id

ChatGPT
Read one instalment One instalment: which schedule it belongs to, its payment number, what was expected and when, what actually landed, the variance, any penalty, the method, the ACH trace and external transaction references, whether it has been reconciled, and why it failed if it did. Money in exact minor units. For the rows as a PAGE — the whole Payments board with its columns, saved views and facet filters, or one deal's rows through ?dealRowId= — read get_definitions_key_rows with key payments; this route is the servicing operation itself. GET /v1/payments/:id (access: workspace)

get_payments_upcoming

ChatGPT
The instalments still to come The instalments still scheduled from today onwards, soonest first — what is due to be collected and when. ?through= is an RFC3339 end of the window, ?limit= caps the page (100 by default, 500 max). Reach for it to answer what is coming in this week or this month; get_payments_overdue has what is already late, and get_portfolio_projection answers the same question as weekly totals for the next quarter rather than row by row. GET /v1/payments/upcoming (access: workspace)

get_pipeline_snapshot

ChatGPT
ONE-CALL pipeline briefing scoped to the authenticated user's data access: deal counts per lifecycle stage, SLA-at-risk deals, closing-soon deals (decision/funding), and stalled deals with no stage movement in stale_days days. Lists are capped (top 10 per bucket) — it is a briefing, not a report. PREFER THIS over multiple search_records calls when the user asks how the pipeline looks, what needs attention today, or for a digest. Read-only.

get_portfolio_aging

ChatGPT
Portfolio aging One row per payment schedule, worst arrears first: what is outstanding, how many days delinquent, and the overdue money split into the same four aging bands. Filter with status= and overdue=true, page with limit. Use it after portfolio health to see WHICH facilities are behind rather than how much is. GET /v1/portfolio/aging (access: workspace)

get_portfolio_health

ChatGPT
Portfolio health The funded book right now: schedules out, collected, outstanding and what is overdue split into 1-30 / 31-60 / 61-90 / 90+ day bands, per currency, plus concentration by lender and by industry. The money figures come from the servicing ledger in exact minor units; the concentration comes from the deals board's own cells, so the two are never summed. Reach for it to answer how big the book is, how much is late, and who it is concentrated with. GET /v1/portfolio/health (access: workspace)

get_portfolio_projection

ChatGPT
Collections projection Expected collections week by week for the next 13 weeks, read from the scheduled instalments the servicing ledger holds (never from cash flows, which can be fees or outbound disbursements). Every week in the horizon is present, quiet ones as zero. Use it to answer what is due to come in over the coming quarter. GET /v1/portfolio/projection (access: workspace)

get_preferences

ChatGPT
Read preferences The signed-in person's OWN app preferences — the desktop's saved arrangements and settings, each a namespaced key (setting:theme, table:leads) over a JSON value this service stores opaquely and never reads. ?prefix= narrows to one namespace. Only the caller's rows are reachable; the workspace-wide defaults under the same keys are get_workspace_preferences, and a grid's typed column layout is get_table-preferences. GET /v1/preferences (access: workspace)

get_preferences_key

ChatGPT
Read preferences key One of the signed-in person's own app preferences by its namespaced key (setting:theme, table:leads) — the stored JSON handed back exactly as it was written, because this service never interprets it. get_preferences lists them and put_preferences_key writes one. A table's column order is deliberately NOT here: that is typed, and lives at get_table-preferences_tableKey. GET /v1/preferences/:key (access: workspace)

get_public_application_form

ChatGPT
Fetches the public signed intake form for a merchant or broker without requiring a Specter user account. Use when a non-user wants to apply for business financing or a broker wants to start signup. It returns the live form schema and public bank-login/upload configuration; it does not submit anything.

get_reports_catalogue

ChatGPT
What a report can be built from The vocabulary behind the Reports page: every source a report can count (deals, leads, submissions, offers, tasks) with its fields and what each field is (dimension, measure, date or computed KPI) and its unit, the aggregates (count, sum, avg, min, max), the operators, the periods a date can be grouped by, and max_rows. Read this before writing a definition for post_reports_preview or post_reports_views — a field that is not here cannot be asked for, and a figure this schema cannot answer is absent rather than zero. GET /v1/reports/catalogue (access: workspace)

get_reports_views

ChatGPT
Side effects: Seeds the workspace's initial report views when absent. The saved reports The reports saved on the Reports page — the caller's own plus the ones shared with the workspace, built-ins first — with their names, descriptions and definitions. NOT the numbers: nothing is cached, so the figures come from post_reports_views_id_run. A fresh workspace's four built-in reports are seeded on the first look, so this is never empty. GET /v1/reports/views (access: workspace)

get_reports_views_id_shares

ChatGPT
Public links on one report The public links published from one report — live ones AND revoked or expired ones — so the workspace can see what has been put on a URL a stranger can open, including links a colleague published. Never a token: that exists in the publish response once and only its hash is kept. GET /v1/reports/views/:id/shares (access: workspace)

get_risk-profiles

ChatGPT
The score bands offers are priced against The risk profiles the offer generator bands a deal by. Global reference data: every seat reads the same rows and no route writes them — growing the vocabulary is a schema change. GET /v1/risk-profiles (access: workspace)

get_servicing_delinquencies

ChatGPT
The deals that are behind The collections cases open against deals that have fallen behind, longest-open first: how many days delinquent, how many payments were missed, the missed amount and the total outstanding, the severity band (early, moderate, severe, critical), how many times the borrower has been contacted and how the last attempt went. ?status= narrows to active, escalated, cured, written_off or settled; ?limit= caps the page. This is the CASE and its chasing trail; get_payments_overdue is the late instalments underneath it, and post_servicing_delinquencies_sweep opens and cures these rows from the ledger in one pass. GET /v1/servicing/delinquencies (access: workspace)

get_servicing_outcomes

ChatGPT
How facilities ended How this workspace's funded deals finished — paid in full, an early payoff, default, a write-off, a settlement, refinanced, renewed, charged off, bankruptcy — newest first, each with what was funded, the payback expected, what was collected, what is still outstanding, what was written off, fees and penalties, the net profit or loss, the collection rate, the settlement amount and discount, days to outcome and total days delinquent. ?outcomeType= narrows it, ?rowId= narrows it to one record's ending (none until that deal ends) and ?limit= caps the page. Reach for it for closed-book questions — recovery, write-offs, payoffs; the LIVE book is get_portfolio_health and get_payment_schedules. GET /v1/servicing/outcomes (access: workspace)

get_signing-requests_id

ChatGPT
Read a signing request One signing request and where each of its signers has got to. Reach for this to check a single envelope after a send or a reminder: it carries the status (open, completed, cancelled, expired), the expiry, and per signer their role, position, whether they have viewed, signed or declined, and how many times they have been chased. GET /v1/signing-requests/:id (access: workspace)

get_signing-requests_id_executed

ChatGPT
Read the executed document The executed document of a completed signing request: self-contained HTML carrying the words exactly as they were signed, every signature image inline, and the audit trail (who, when, from what address, with what consent). Reach for this to hand somebody the finished paper. It is a not-found while any signer is still outstanding. GET /v1/signing-requests/:id/executed (access: workspace)

get_sms-blasts_id

ChatGPT
Read one text blast One TEXT-message blast by id: its words, status, audience size, the outcome counts and a sample of up to 25 recipients. The kind comes from the path, so an email blast's id answers not-found here (its twin is get_email-blasts_id). This is the cheap read; get_blasts_id_report is the whole picture of one blast — every recipient, the thread behind each address and whether they replied. GET /v1/sms-blasts/:id (access: workspace)

get_sms_blast

ChatGPT
For non-merchant Specter broker/operator users only. Reads one SMS blast in full: what was sent, to how many, and how each recipient landed. Use to answer "how did that blast do" before deciding whether to retarget it. Read-only.

get_sms_sender_access

ChatGPT
Reports which phone numbers the authenticated user may send SMS from. Use before sending SMS through MCP: send_channel_message and send_conversation_message accept sender_phone, and the backend rejects a number the user has no access to. This returns an ACCESS MAP, not a complete list of numbers — read it as an exclusion list. blocked_numbers are the numbers this user may NOT send from (assigned to another user, or a Zoom line with no assignee). allowed_numbers are numbers explicitly assigned to this user. A number restricted to nobody in particular is unrestricted and appears in NEITHER list, and when is_admin is true both lists come back empty because the whole tenant pool is available. So an empty allowed_numbers never means the user cannot send — do not refuse on that basis. To name a specific number, take one from allowed_numbers; otherwise send without sender_phone and let the backend choose, and only warn the user if the number you intended appears in blocked_numbers. It is read-only.

get_storage-accounts

ChatGPT
Read storage-accounts Where this workspace's documents are stored and under whose credentials: every connected file store with its provider, its label, the connected account it rides on and the root folder new files land in. Any seat may read it; connecting one (post_storage-accounts) and detaching one are a workspace admin's. GET /v1/storage-accounts (access: workspace)

get_submission_history

ChatGPT
Reads timeline/history for a lender submission visible to the authenticated user. Use when the user asks for submission history, status changes, lender events, or related activity. It is read-only.

get_submissions

ChatGPT
List the deals sent to lenders through Specter Sends made in Specter, one per deal-to-lender crossing — NOT the Submissions page, whose imported history is not here. To count or list submissions, read that page (get_definitions_key_rows_facets, params.key = "submissions"). GET /v1/submissions (access: workspace)

get_submissions_analytics

ChatGPT
Lender decision rates for deals sent through Specter Approval rate, decline rate, average hours to a lender's decision and average offer over a window, with a breakdown row per lender. WHAT IT COUNTS: only the crossings get_submissions lists — sends made in Specter, which stamp when they went out and when the lender decided. A submission recorded on the Submissions page without a send in Specter (imported history) carries no such stamp and is not counted, so on a workspace whose history was imported this counts a few crossings while the page holds thousands. source=page counts the Submissions page's rows instead, imported history included: every lender ranked by submissions with its approved, closed and funded counts and approval rate, the whole page unless a window is passed (then the rows sent in it) — the one call for "top lenders by approval rate". Rates are absent (null) rather than zero when nothing was decided. Pass window=30d|90d|180d|365d. One at a time: get_lenders_id_analytics GET /v1/submissions/analytics (access: workspace)

get_table-preferences

ChatGPT
Read table-preferences Every grid layout the signed-in person has saved, keyed by table (table:deals:overview and its like) — column order, hidden columns, the sort and the grouping. Each member's own arrangements and nobody else's. Column WIDTHS are not here: a width belongs to the definition's column rather than to a person. GET /v1/table-preferences (access: workspace)

get_table-preferences_tableKey

ChatGPT
Read table-preferences tableKey One grid's layout as the signed-in person arranged it — tableKey namespaced like table:deals:overview — with its column order, hidden columns, sort column and direction, and grouping column. Typed, unlike the opaque blob get_preferences_key answers, so the column ids in it can be reasoned about. put_table-preferences_tableKey replaces it whole. GET /v1/table-preferences/:tableKey (access: workspace)

get_task

ChatGPT
For non-merchant Specter broker/operator users only. Reads one task in full by its ID, after a list tool surfaced it. Read-only.

get_task_stats

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL task posture for the tenant: totals by status, SLA at-risk and breached counts, due today, breakdowns by type and priority, completion rate and average completion time. Takes no arguments. Use for "how are we doing on tasks" instead of paging a list and counting. Read-only.

get_tasks

ChatGPT
List tasks The workspace's tasks: what is on whose plate, and by when. A task on a record carries record: {key,rowId,label} — the page, the row and the record's name; join a task to its deal or lead by record.rowId, never by its title. label is null when the person may not open that record. summary=true answers each as {id,title,status,dueAt,assignee,entity} — the assignee by name, the record by {kind,key,rowId,label}. GET /v1/tasks (access: workspace)

get_tasks_id_comments

ChatGPT
Read a task's comments What has been written on one task, oldest first — people's comments and the system's own notes of what changed. limit defaults to 100 and is capped at 500. GET /v1/tasks/:id/comments (access: workspace)

get_tasks_stats

ChatGPT
Read task figures How the workspace's queue stands: how many tasks are open, late, due today, breaching SLA, and the open ones broken down by type and by priority. Counts the WHOLE queue, so reach for this instead of adding up a page of GET /v1/tasks. GET /v1/tasks/stats (access: workspace)

get_tax_return

ChatGPT
Gets one tax return visible to the authenticated user. Use when the user asks for tax return details, years, profitability summary, linked application, or risk signals. It is read-only.

get_tax_return_history

ChatGPT
Reads timeline/history for a tax return visible to the authenticated user. Use when the user asks how a tax return or its analysis changed. It is read-only.

get_tax_return_year

ChatGPT
Gets one tax return year visible to the authenticated user. Use for annual tax metrics, extracted fields, profitability, ownership, and document URLs. It is read-only.

get_thread_for_entity

ChatGPT
For non-merchant Specter broker/operator users only. Finds the INTERNAL collaboration thread attached to a record — where colleagues discuss a deal among themselves. This is NOT merchant messaging: outbound and inbound email/SMS with the merchant live on communication channels, via get_communication_channel_for_entity. Returns exactly one thread, or a not-found error when the record has none, which means nobody has started an internal discussion — not that the record is missing. Use it to get the thread_id that list_thread_messages needs. Read-only.

get_threads

ChatGPT
List discussion threads The workspace's INTERNAL threads — the Internal section of Communications: every discussion colleagues have started on a record, whatever page that record sits on. ?status= narrows by state, ?mine=true to the threads the caller is a participant in, ?limit= caps it. The one read in this family that is not record-scoped; for one record's threads read get_definitions_key_rows_rowId_threads. These are people here talking to each other about a deal, not messages to a merchant — those are the comms conversations. GET /v1/threads (access: workspace)

get_underwriting_packet

ChatGPT
ONE-CALL briefing for a single facility/application: the enriched deal (revenue, credit, health score, AI summary, assignments) PLUS its lender submissions, offers, bank statements with risk signals, and recent timeline. PREFER THIS over separate get_application_details / list_application_submissions / list_application_bank_statements / get_application_history calls whenever the user asks broadly about one deal ('tell me about', 'summarize', 'what's the status of', /lookup, /analyze prep) — it replaces 4-5 round trips with one. Read-only. Sections may carry a section_errors entry instead of failing the whole packet.

get_undo_history

ChatGPT
Read my undo history The signed-in person's own recent changes, newest first, as Command-Z sees them: each operation (one request, a whole X-Undo-Group bulk action, or one Neo turn) with its sentence ("Converted 3 leads to deals", "Edited a record on Leads"), who made it (person or neo), how many rows it touched, and undoneAt once taken back. limit 1-100, default 20. Read it before post.undo.last when the person asks what an undo would take back, or names an older change to undo by its id. Never a colleague's: the seat comes from the session. GET /v1/undo/history (access: workspace)

get_webhooks

ChatGPT
Read webhooks Every outbound subscription this workspace has — endpoint, events, status, auth kind and the signing secret's hint only — plus the two closed vocabularies (the events that can be subscribed to, and the credential kinds a receiver can ask for) so a caller offers exactly what the database will accept. Admin only, reading included. GET /v1/webhooks (access: workspace:admin)

get_webhooks_id

ChatGPT
Read webhooks id One subscription: its endpoint, events, status, auth kind and secret hint. The signing secret itself is NOT here — only the create and the rotate answers ever carry it. Admin only. GET /v1/webhooks/:id (access: workspace:admin)

get_webhooks_id_deliveries

ChatGPT
Read webhooks id deliveries The delivery attempts for one subscription, newest first: event, attempt number, outcome, response status, error text and duration. This is where a test press or a failing endpoint is actually diagnosed. limit defaults to 50, and anything over 200 falls back to 50. Admin only. GET /v1/webhooks/:id/deliveries (access: workspace:admin)

get_workflow

ChatGPT
Gets one Workflow V3 workflow by ID, including its full trigger, nodes, and edges. Use before editing, auditing, re-validating, or publishing an EXISTING workflow when its graph is not already in this conversation — update_workflow_draft and validate_workflow_draft need the full graph, and this is the only way to read it back. Read-only. The model-visible response is size-capped: if any array carries a truncated marker or the body itself is truncated, the graph is INCOMPLETE — never send it back through update_workflow_draft (that call replaces nodes and edges wholesale and would delete the missing ones); tell the user to make that edit in the builder instead.

get_workflow_builder_tenant_capabilities

ChatGPT
Reads tenant-scoped Workflow V3 readiness for communication nodes, including Zoom SMS sender phones, Twilio SMS configuration status, and the email sender validation path. Use before configuring email/SMS actions so the assistant does not invent sender emails, sender phones, provider IDs, or integration settings. It never returns provider secrets.

get_workflow_fos_entity_fields

ChatGPT
Reads the field schema for one FOS entity type available to Workflow V3. Use before configuring conditions, mappings, or node data fields. It is read-only.

get_workflows

ChatGPT
List workflows Every automation workflow in this workspace with its trigger and its steps, newest first — the rules that fire on their own. summary=true answers each without its steps: its trigger, stepCount and lastRun. It answers the workspace's OWN workflows only; for the vocabulary a workflow may be built FROM (the trigger kinds, step kinds, fields and operators) reach for get_workflows_catalog, which is the same for every workspace and is what a builder needs before anything has been created. GET /v1/workflows (access: workspace)

get_workflows_catalog

ChatGPT
The workflow vocabulary Everything a workflow may be BUILT out of: every action a step can call, every event a trigger can listen for, the fields those events carry, the presets, the step kinds, the trigger kinds (event, schedule, manual) and the comparison operators. Read this before writing or validating a definition — a step naming an action that is not here is exactly what post_workflows_validate refuses. Whole it is megabytes: summary=true answers everything but the actions, each event as its key and entity kind — a step's action is any tool's key, and its tool name names it too; keys= (comma list of action keys, tool names or event keys) answers just those whole, and q= the actions and events whose key, title or description holds the term. It is the vocabulary, not a workspace's data, so every seat sees the same one; for the rules this workspace has actually written, get_workflows. A step naming a preset puts the preset's own fields straight in with; a step naming an action puts its arguments under with's params, query and body. GET /v1/workflows/catalog (access: workspace)

get_workflows_id

ChatGPT
One workflow and its draft's problems One automation rule in full — its trigger, its steps, its status — plus the problems its CURRENT DRAFT has, so a builder can flag the canvas the moment it opens. Those problems are the same ones a publish refuses on. Reading changes nothing; the frozen words a run is actually executing are in get_workflows_id_versions. GET /v1/workflows/:id (access: workspace)

get_workflows_id_versions

ChatGPT
Every publish of one workflow Every version one workflow has published, newest first: what each publish froze, when, and which version a rollback republished. A version is immutable and a run points at one, so this is the history that reads FORWARDS — a rollback publishes an old version's words as a new version rather than restoring it. The draft being edited is in get_workflows_id. GET /v1/workflows/:id/versions (access: workspace)

get_workflows_reviews

ChatGPT
Workflow steps waiting on a person The automation steps that stopped to ask somebody — what is on whose plate from a workflow, and what it is waiting to be told. ?mine=1 narrows it to the caller: reviews addressed to them by name plus every review addressed to a role they hold. Answering one, which lets the run carry on, is post_workflows_reviews_id_decide. GET /v1/workflows/reviews (access: workspace)

get_workflows_runs

ChatGPT
Workflow runs, newest first The record of automations actually executing: one entry per run with its workflow, what triggered it, its status and where it got to — "did that rule fire last night". Filters: ?workflowId=, ?status=, ?limit=; summary=true leaves out each run's context. For one run's step-by-step story and why it stopped, get_workflows_runs_id; for the rules themselves rather than their executions, get_workflows. GET /v1/workflows/runs (access: workspace)

get_workflows_runs_id

ChatGPT
One run, its steps and its reviews One workflow run in full: the run, every step and what became of it, and any review it asked for — all three together, because "why did this stop" is one question. This is the read that answers a failed or stuck automation. It changes nothing; stopping a run that is still going is post_workflows_runs_id_cancel. GET /v1/workflows/runs/:id (access: workspace)

get_workspace_access-denials

ChatGPT
Read the workspace's refused calls Admin only. Every 403 this service has answered for this workspace, newest first: the route template, the access class that route requires, the seat the caller actually held, the service's own reason, and the request id so the refusal and its log line can be read together. Paged with page and limit, 50 by default. Two absences are deliberate - a refusal answered before a workspace was established has no workspace feed to land in, and a 401 is not a refusal, it is not being signed in. GET /v1/workspace/access-denials (access: workspace:admin)

get_workspace_agreements_kind

ChatGPT
Read workspace agreements kind One agreement addressed by KIND ('msa'), with this workspace's standing against the current version: whether the caller personally accepted it, whether the workspace signed and who signed, and when this version was first and last viewed. The HTML body is present only when the words are held here rather than in the CRM. An unknown kind is a 404. Any seat. GET /v1/workspace/agreements/:kind (access: workspace)

get_workspace_automated-emails

ChatGPT
Read the workspace's automated emails Which system mails this service still sends about the workspace's deals — facility status, the four submission notices, SBA readiness and the merchant decline notice — with each one's subject prefix and introduction, plus the sender name and reply-to they go out under. Every kind is always present, so this is the list of what a workspace can switch off as well as what it has. Never 404s: an untouched workspace reads the defaults, which are all on. GET /v1/workspace/automated-emails (access: workspace)

get_workspace_compliance_reports

ChatGPT
Read the workspace's compliance reports Admin only. The standing questions this workspace asks of itself, each with its newest run and where that run stands, plus the closed vocabulary of checks and what to print for each. The checks are: every funded deal has an executed contract, every shared submission has a disclosure on file, every member's seat has been reviewed, and every active vendor's contract is current. GET /v1/workspace/compliance/reports (access: workspace:admin)

get_workspace_compliance_reports_id_runs

ChatGPT
Read a report's runs Admin only. One report's history of asking, newest first, each with its outcome, its counts and the trail of decisions people recorded on it. limit defaults to 25. GET /v1/workspace/compliance/reports/:id/runs (access: workspace:admin)

get_workspace_compliance_runs_runId

ChatGPT
Read one compliance run Admin only. The run, everything it saw, and the trail of what people concluded. Each finding names the check, the subject (a deal, a submission, a member, a vendor or the workspace itself), whether it passed, and the reason in the check's own words. This is also the export: nothing is cached, so a client writes its CSV from this answer rather than from a second, staler copy. GET /v1/workspace/compliance/runs/:runId (access: workspace:admin)

get_workspace_delegation

ChatGPT
Read workspace delegation Which domains this workspace has delegated to us, per provider, each with its status and how many inboxes the grant has minted — plus the public half a domain administrator needs: our Google service account's client id and the scopes to authorise, and the Microsoft admin-consent URL. THE CREDENTIAL IS NEVER IN THIS ANSWER, and a deployment that cannot do Google delegation answers an empty client id rather than inviting consent for a key we do not hold. Any seat and not plan-gated, unlike every write beside it. GET /v1/workspace/delegation (access: workspace)

get_workspace_directory

ChatGPT
List the connected directory's people Everyone in the Google Workspace or Microsoft 365 domain this workspace has a verified delegation for, each placed against the roster: seated (already has a seat here, with their account id and seat), invited (an open invite already names their address) or new. Each row also says whether the provider has suspended the account. Reach for it to answer 'who at the company still has no access' or 'has anyone who left still got a seat'. Needs a connected delegation; admin and Enterprise. GET /v1/workspace/directory (access: workspace:admin)

get_workspace_disclosure-states

ChatGPT
Read the states this workspace discloses in The seven US states with a commercial financing disclosure law (CA, NY, UT, VA, GA, FL, CT) plus any this workspace added, each saying whether the law wants the merchant's signature. Reach for it to decide whether a submission needs a disclosure before it goes to a lender. GET /v1/workspace/disclosure-states (access: workspace)

get_workspace_document-types

ChatGPT
List the workspace's document types Every document type a file, a document request, a record's checklist standing, a checklist rule, a result mapping or a message template may name: builtIn, the app's own words each with its label (bank_statement as Bank statement, …), and custom, the workspace's own processor-v2 types each as {word, name} (word is custom_<slug>, the value to send). customAvailable is false, with note saying why, when this deployment has no processor-v2 — then only the built-ins exist. A custom word not in this list is refused 422 document_type_unknown by every write. Any seat. GET /v1/workspace/document-types (access: workspace)

get_workspace_domains

ChatGPT
List workspace domains Every domain this workspace claims to send mail as, each with its DNS records, whether each record is currently in place, its status (pending, verified or failing) and how many inbound addresses receive at it. Reach for it to answer 'can we send as our own domain yet' or 'why did our mail stop being delivered'. GET /v1/workspace/domains (access: workspace)

get_workspace_domains_id

ChatGPT
Read one workspace domain One claimed domain with its full record list and the result of the last DNS check per record, plus formHost - the address under it the workspace's form links use, its CNAME record and status - or null. Reach for it when you need the exact host and value a person has to paste into their DNS. GET /v1/workspace/domains/:id (access: workspace)

get_workspace_domains_id_inbound

ChatGPT
List a domain's inbound addresses The addresses at one domain that mail may arrive at, each with what an arriving message opens (a pack page, a board, or a member's inbox) and whether it is still receiving. GET /v1/workspace/domains/:id/inbound (access: workspace)

get_workspace_drive-folders

ChatGPT
Browse Drive folders The folders directly inside ?parent=, or the connected Google account's My Drive root when parent is absent - the folder picker's read, for choosing what to watch. Folders only: it cannot list a document and cannot be widened into a search. Answers the account whose Drive was read, so a picker can say whose it is. GET /v1/workspace/drive-folders (access: workspace:admin)

get_workspace_drive-watches

ChatGPT
List Drive watches Every Google Drive and OneDrive folder this workspace watches (provider google_drive or onedrive), with where each one lands its new files (a pack page row, a board row, or the workspace document library), when it was last swept, how many documents that sweep landed and how many it has landed in total. Also answers the closed lists of target and provider words and whether a sweep can actually run on this deployment, overall and per provider. Admin only. Reach for it before adding a watch, to see whether the folder is already watched. GET /v1/workspace/drive-watches (access: workspace:admin)

get_workspace_execution-profile

ChatGPT
Read the execution profile Who signs for this workspace by default - the name, officer title and address that go on its own papers. Reach for this before sending a document the company counter-signs; configured is false when nobody has been named, and the send then refuses rather than signing under a blank name. GET /v1/workspace/execution-profile (access: workspace)

get_workspace_heartbeat

ChatGPT
Read workspace heartbeat What the workspace's autonomous heartbeat is allowed to do: whether it is on, its model and reasoning effort, which events wake it, which action kinds it may plan, which of those run on acceptance, its minimum wake interval and per-run action ceiling, the follow-up rules and the queue's lifecycle bounds. A workspace that has never opened the tab reads the defaults, which are OFF, rather than a 404. ANY SEAT — what a machine does inside a workspace is not a secret from the people it does it to. GET /v1/workspace/heartbeat (access: workspace)

get_workspace_heartbeat_actions

ChatGPT
Read workspace heartbeat actions Read due, pending follow-up drafts awaiting a person from live runs across the workspace, restricted to conversations visible to the caller. Use for the Notifications decision inbox. Pages newest first using before and limit (1–100); next_before is zero at the end. can_decide identifies whether this seat may use the existing admin decision routes. Dry runs, settled actions and future deferrals are excluded. GET /v1/workspace/heartbeat/actions (access: workspace)

get_workspace_heartbeat_runs

ChatGPT
List the heartbeat's runs The heartbeat's recent runs, newest first — when it woke, what woke it and what it planned — together with the closed vocabularies a settings pane needs: the action kinds, the wake events, the models and the effort levels. limit defaults to 50 and anything over 200 falls back to 50. Any seat. GET /v1/workspace/heartbeat/runs (access: workspace)

get_workspace_heartbeat_stops

ChatGPT
List conversations the heartbeat is stopped for The conversations somebody told the heartbeat to stop drafting for — contact, medium and when — limited to those the caller can read. Reach for it to answer "why is nothing being drafted for this merchant" or to find one to resume. A stop also covers every other conversation filed under the same record, which this list does not repeat. GET /v1/workspace/heartbeat/stops (access: workspace)

get_workspace_invites

ChatGPT
List workspace invitations and their status Every invite link this workspace has minted: its role, who made it, when it expires, and whether it was accepted or revoked. The token is not here and can never be read again — a lost link is re-minted, not recovered. Admin only. GET /v1/workspace/invites (access: workspace:admin)

get_workspace_lender-catalogue

ChatGPT
List the published lender catalogue The lenders the platform publishes, each with its products, states served and funding range, and — where this workspace has already adopted one — the id of the record it made on their own Lenders board. Admin only. Read it to see what could be adopted and what already was. GET /v1/workspace/lender-catalogue (access: workspace:admin)

get_workspace_mcp_usage

ChatGPT
Read workspace mcp usage What the models connected to this workspace have been doing: which MCP tools, from which clients, driven by which person or which key, folded per day, with the failures. tool narrows inside MCP traffic rather than matching any route that shares the word; key, client and class narrow further. ADMIN ONLY and deliberately NOT plan-gated — a workspace whose Workspace API tier lapsed still needs to see what its models did while it had one. What the PEOPLE did is /v1/workspace/usage/summary. GET /v1/workspace/mcp/usage (access: workspace:admin)

get_workspace_members

ChatGPT
List the workspace's members and their roles The workspace's roster: every seat with the person behind it and their role. ANY SEAT may read it — a board maker below admin needs the list to grant access and assign rows. Changing a role is the admin-only PATCH beside it. GET /v1/workspace/members (access: workspace)

get_workspace_onedrive-folders

ChatGPT
Browse OneDrive folders The folders directly inside ?parent= of the caller's own OneDrive, or of the SharePoint library ?drive= names (an id from GET /v1/onedrive/libraries); the top folder when parent is absent. The OneDrive arm of the drive-watch folder picker, read with the caller's own Microsoft connection - the one a watch they add will sweep with. Folders only: it cannot list a document. Admin only. GET /v1/workspace/onedrive-folders (access: workspace:admin)

get_workspace_overlaps

ChatGPT
List businesses that appear on more than one page WHAT LOOKS WRONG in the records: the parties this workspace holds TWICE on one page (duplicate) and the ones sitting on TWO different pages at once (two_places, e.g. the same business as a lead and as a deal). Matched three ways — the business name with case, punctuation and LLC/Inc/Corp stripped, the email address, the phone number — and each group names the shared value plus every record holding it, with its page and row id so you can open them. Reach for it UNASKED when somebody asks what needs attention, what to clean up, whether a record already exists before making another, or why two people are working the same account; and reach for it before creating a lead, deal or merchant for a business whose name you have just been given. NOTHING IS CHANGED by this call and there is no merge: it names the pairs and the person decides. A lead that was converted into a deal is NOT reported, nor are records already linked to each other by a relation cell, nor the contacts minted from inbound mail — those pairs are the way the desk is supposed to work. scanned is how many records were compared and truncated says the answer was cut, so an empty report over a truncated scan does not mean nothing is wrong. Submissions, offers and payments are not read: their identity is a reference number and their contact columns hold the other party's address. GET /v1/workspace/overlaps (access: workspace)

get_workspace_plan

ChatGPT
Read the workspace's plan and seats What the workspace is ENTITLED to: its tier, business type, subscription status, trial end, when it subscribed, and how many seats it holds. A workspace with no plan row reads as the entry tier rather than an error, and a trial that has run out reads as the entry tier too. Any seat. This is the entitlement — /v1/billing is the money — so a caller that only needs the tier should read this one. GET /v1/workspace/plan (access: workspace)

get_workspace_portal-access

ChatGPT
How long client portal access lasts The workspace's rule for how long a borrower's (or outside affiliate's) CLAIMED portal access lasts: while_open (default — until the record reaches a status its page marks terminal), link_expiry (until the invitation link's own expiry), or days_after_close with days 1–365. Revoking an invitation always ends access at once. Admin seat. GET /v1/workspace/portal-access (access: workspace:admin)

get_workspace_preferences

ChatGPT
Read workspace preferences The workspace's SHARED defaults — what an admin published for everyone, under namespaced keys like 'table:leads' or 'setting:theme'. prefix narrows by the start of the key. Any seat reads it, because a default nobody may read cannot be applied. A member's own overrides are /v1/preferences. GET /v1/workspace/preferences (access: workspace)

get_workspace_preferences_key

ChatGPT
Read workspace preferences key One shared default by its namespaced key. A key nobody has published is a not-found, which is how a caller distinguishes 'this workspace has no default' from 'the default is null'. Any seat. GET /v1/workspace/preferences/:key (access: workspace)

get_workspace_processor

ChatGPT
Read where the workspace's applications arrive The address this workspace's applications arrive at (empty means the platform's shared address), each Gmail account the workspace connected for it with its health and aliases, what the workspace's own domains have set up, and whether an alias can be created at Google through its domain-wide delegation. Reach for it to answer 'where do our applications go'. GET /v1/workspace/processor (access: workspace)

get_workspace_processor_connections_mailbox_send-as

ChatGPT
List an intake mailbox's existing Gmail aliases The addresses Gmail already delivers to one of the workspace's processor accounts (its send-as list), read with the account's own grant, so one can be picked as an alias without creating anything. GET /v1/workspace/processor/connections/:mailbox/send-as (access: workspace:admin)

get_workspace_processor_criteria

ChatGPT
Read what counts as a processed deal This workspace's processor@ criteria: the intake rules that decide whether an arriving email is taken at all ({signal, comparison, value, onFail} over its files' count, kinds, sizes, locks, duplicates and classified types), the completeness rules per product that decide missing information versus a complete deal (what the documents contain, each required, nice_to_have or warning, in plain words), whether a deal is made on arrival or only once no required rule fails, which emails go out, the deals-board status a new deal lands in and its default assignee — plus the presets the pane can apply. isDefault is true while the workspace never saved any. Reach for it to explain why an emailed application was held or refused. Any seat. GET /v1/workspace/processor/criteria (access: workspace)

get_workspace_processor_file-types_key_examples

ChatGPT
Read a file type's examples The documents a person gave processor-v2 as examples of one file type, each with its id (the document's sha256), file name, status (reading while processor-v2 reads it, learned once read, failed when nothing could be read) and the clues it suggests the type lacks. :key is the app's document word (bank_statement, id, …) or a workspace's own custom_<slug> type; 400 for a word processor-v2 has no type for, 503 when this deployment has no processor-v2. Any seat. GET /v1/workspace/processor/file-types/:key/examples (access: workspace)

get_workspace_processor_result-mappings

ChatGPT
Read where file types' results go The workspace's result mappings: for each file type (documentType, the app's document word or a custom_<slug> type) and board (definitionId), which extracted field goes to which column — on the same row (target same_row) or on the row a relation column links to (linked_row, relationColumnId) — narrowed to one file column when fileColumnId is set, with its updateRule (fill_empty writes only into an empty cell, review holds the value for a person, overwrite replaces a value unless another document wrote it). status is active, or needs_attention with problem column_missing, column_changed, relation_missing or file_column_missing when the board changed since; such a mapping writes nothing. ?documentType= narrows to one file type. Any seat. GET /v1/workspace/processor/result-mappings (access: workspace)

get_workspace_processor_whitelist

ChatGPT
Read the workspace's processor whitelist The senders with no Specter seat this workspace lets email processor@ on its behalf — whole domains or single addresses — each with why and who added them. A matching sender's email files into THIS workspace as the seat of the admin who added the entry; an address entry beats a domain entry. Any seat. GET /v1/workspace/processor/whitelist (access: workspace)

get_workspace_readiness

ChatGPT
Read what the workspace still needs set up The paperwork between this workspace and the Lenders surface, answered PER CALLER: the workspace's KYC status, the workspace's signature on the current MSA, and whether THIS person has accepted it. A KYC that is submitted but not yet verified already counts as unlocked; in_progress and rejected do not. Any seat. GET /v1/workspace/readiness (access: workspace)

get_workspace_rep

ChatGPT
Read workspace rep Who looks after this workspace: the assigned representative's name, title, email, published phone (blank when we publish none) and thread id, which is 0 until the thread has been opened. 404 with a sentence when nobody is assigned — never a card naming somebody who does not work here — and the sentence does not say why. Any seat. GET /v1/workspace/rep (access: workspace)

get_workspace_rep_thread

ChatGPT
Side effects: Marks the conversation and related notifications read. Read workspace rep thread The conversation with this workspace's representative, most recent first under limit. A rep assigned but never written to answers an empty list; no rep at all is the same 404 the card gives. Any seat. GET /v1/workspace/rep/thread (access: workspace)

get_workspace_role-presets

ChatGPT
List the role presets a pack installed Every ROLE PRESET held here: the roles a pack shipped (the law packs ship Managing partner, Attorney, Associate attorney, Paralegal, Case manager, Intake specialist, Legal assistant, Bookkeeper, and the outside affiliates Co-counsel, Medical provider and Investigator) and the ones the firm made from them. A STAFF preset is a seat (member, team_lead or admin, never owner), a level per board of its pack (none, read, write) and a books level (read, manage or none); scope says which rows the seat sees (assigned for a member, all for a team lead or admin). An AFFILIATE preset is never a seat: affiliate lists the boards an outsider may be invited onto, the fields they read, the case identity they are shown, whether they see only their own uploads, and the document types they may file. Any seat may read it. GET /v1/workspace/role-presets (access: workspace)

get_workspace_sidebar-folders

ChatGPT
Read the sidebar folders The workspace's sidebar folders in rail order, each with its filed items in order. Folders are shared with every seat. An item is a page's route word ("deals") or a board as "board:<board id>"; a board item is left out for a caller who cannot open that board, so a folder never reveals a board you were not given. Route items are answered as stored. Any seat. GET /v1/workspace/sidebar-folders (access: workspace)

get_workspace_sla-policies

ChatGPT
Read the workspace's SLA policies The promises this desk makes: per pack page and per stage, how long a record may sit before it is at risk and before it has breached, and the calendar the clock runs on (business hours, time zone, working days). Any seat may read them. Targets are in MINUTES, so a four-hour promise is expressible; a stage with no policy has no promise, and its records answer state "none" rather than "on track". GET /v1/workspace/sla-policies (access: workspace)

get_workspace_sla-summary

ChatGPT
Side effects: Sweeps SLA breaches and creates breach notifications before returning the summary. Read the SLA dashboard tile Every live record under a policy, counted on track, at risk and breached, with the same three numbers per stage. Nothing is cached: the dwell is read from the record's own timeline and the state computed on the way out. This is also the ONLY place the breach sweep runs, so reading it records any newly crossed breach and rings its bell exactly once, and the number shown cannot disagree with the bell sent. A summary measuring 0 records is a workspace with no policies, not a quiet one. GET /v1/workspace/sla-summary (access: workspace)

get_workspace_staff-access

ChatGPT
Read workspace staff-access What WE did inside this workspace: every staff reach-in — who, when, why and for how long — and every operator act aimed at it, newest first. Read-only and confined to this workspace's own rows; platform-wide acts aimed at nobody in particular are invisible here by construction. limit defaults to 50 and caps at 200, with offset. Admin only. GET /v1/workspace/staff-access (access: workspace:admin)

get_workspace_summary

ChatGPT
Read the workspace at a glance The workspace at a glance: the pipeline by stage, what is new this week, the comms picture, recent activity and the facility count. The first call for "how are we doing" and "what needs attention". delinquency is how far behind the book is right now: overdue instalments, what they add up to in minor units, and how many schedules carry one — the Payments board's overdue view, counted. GET /v1/workspace/summary (access: workspace)

get_workspace_super-kyc

ChatGPT
Read the workspace's Super KYC settings What this workspace holds a merchant to when Super KYC watches their bank and mail: the smallest deposit from a known lender that counts as an advance (lender_deposit_min_minor) and the smallest funding-shaped deposit from an unknown payee worth raising (unmatched_deposit_min_minor), both in cents; how many NSF or overdraft fees (nsf_count) inside how many days (nsf_window_days) is a warning; the workspace's own words for a lender deposit that is not an advance, added to the built-in ones; the funders it wants matched beside the platform catalogue and the ones it does not want flagged; and the names and domains watched on every link. Never 404s: a workspace that has not saved reads the defaults — $5,000, $10,000, 3 fees in 30 days and four empty lists. ANY SEAT. GET /v1/workspace/super-kyc (access: workspace)

get_workspace_support_tickets

ChatGPT
List support tickets The CALLER'S OWN support tickets, not the workspace's — every read here is scoped to who opened it. Filter by status and priority, window by when they were raised, sort and page at up to 100 a page. Any seat. GET /v1/workspace/support/tickets (access: workspace)

get_workspace_support_tickets_id

ChatGPT
Read a support ticket One of the caller's own tickets with its thread, as the customer sees it: our internal notes on that ticket are excluded here and in the database's own policy. Somebody else's ticket is a not-found. Any seat. GET /v1/workspace/support/tickets/:id (access: workspace)

get_workspace_team-scope

ChatGPT
Read which teams' records the caller sees The account ids whose work the CALLER leads, read from the team rows, including their own when they lead a team. An empty list means they lead no team — which is not 'they see nobody', since a lead with no team still reads the whole workspace. Any seat. GET /v1/workspace/team-scope (access: workspace)

get_workspace_teams

ChatGPT
List the workspace's teams The workspace's teams with their leads and members. Any seat reads it; every write beside it is an admin's. GET /v1/workspace/teams (access: workspace)

get_workspace_underwriting-criteria

ChatGPT
Read the workspace's underwriting criteria The workspace's underwriting scorecard TEMPLATE: the questions this lender asks of every deal, each with its label, category, value kind, threshold and comparison. The ANSWERS live on a record and are served from the deal, not here. Any seat, because the tab has to draw the checklist. GET /v1/workspace/underwriting-criteria (access: workspace)

get_workspace_underwriting_rule-sets

ChatGPT
List the underwriting rule sets Every underwriting rule set this workspace has, each with its DRAFT rules and whether the draft differs from what is published. Reach for it to answer "what rules judge our deals", or before editing one. The rules that actually RUN are the published version's, not these. GET /v1/workspace/underwriting/rule-sets (access: workspace)

get_workspace_underwriting_rule-sets_setId

ChatGPT
Read one underwriting rule set One rule set with its draft rules in evaluation order. Use it to read a rule's condition before changing it. GET /v1/workspace/underwriting/rule-sets/:setId (access: workspace)

get_workspace_underwriting_settings

ChatGPT
Read the underwriting settings The workspace's rules-engine settings. Today one switch: whether a deal is re-evaluated automatically when its typed processing results change. Off unless somebody turned it on. GET /v1/workspace/underwriting/settings (access: workspace)

get_workspace_usage_events

ChatGPT
Read workspace usage events The individual acts behind the usage numbers, newest first and paged: who did what, folded into a feature. Narrow with member (a userId from a summary row) and feature. Reach for it when a usage figure needs explaining — which records, which imports, which messages made up the count. GET /v1/workspace/usage/events (access: workspace:admin)

get_workspace_usage_members

ChatGPT
Read workspace usage per member Per member over a 7d, 30d or 90d window: the desktop installs they launch (platform, version, installed and last seen), the AI apps they connected and each app's MCP calls, their Neo turns, rounds and tokens with their most-used tools, and their MCP calls by client and tool with failures — plus the workspace's line (seats, how many have the desktop, a connected app, use Neo, use MCP). Admin only. Neo's dollar cost is not here. mcpWindow says how much of the window the runs table still holds. GET /v1/workspace/usage/members (access: workspace:admin)

get_workspace_usage_summary

ChatGPT
Read workspace usage summary What this workspace actually uses over a 7d, 30d or 90d window: headline totals (records touched, messages sent, Neo turns, active minutes), a row per member, a row per feature, and a daily series. Answers "who is using this", "which features are used" and "is adoption moving". Admin only; a person's own figures are get.me.usage.summary. GET /v1/workspace/usage/summary (access: workspace:admin)

get_workspace_vendors

ChatGPT
Read the workspace's vendors The outside parties this workspace engages - processors, attorneys, credit bureaux, notaries - with their contacts, contract dates, the files pinned to each, and the closed kind vocabulary a picker needs. Any seat may read: a closer needs to know which attorney is on the file. A vendor with no contract end date on file is a compliance finding rather than a fine state. GET /v1/workspace/vendors (access: workspace)

get_workspace_vocabulary

ChatGPT
Complete a value the way this workspace spells it What this workspace has typed before under a column with this name, as completions for a prefix. header is the column's label — the memory is keyed on what a column MEANS, so Industry on one board answers for industry on another — and prefix is what has been typed so far. Best first, frequency decayed by recency. Reach for it to finish a value the way this workspace already spells it, rather than inventing a new spelling of the same thing. GET /v1/workspace/vocabulary (access: workspace)

list_application_bank_statements

ChatGPT
Lists bank statement records visible to the authenticated user, optionally filtered to one facility/application, counterparty, or status. Use when the user asks for statements, statement files, processing state, or statement analysis linked to an application. It is read-only.

list_application_submissions

ChatGPT
Lists lender submissions visible to the authenticated user, optionally filtered to one facility/application, lender, status, or recent-day window. Use when the user asks where an application was submitted, recent lender activity, or submission status. It is read-only.

list_application_tax_returns

ChatGPT
Lists tax return records visible to the authenticated user, optionally filtered to one facility/application, counterparty, or status. Use when the user asks for tax returns, profitability, tax years, or tax-document analysis linked to an application. It is read-only.

list_available_email_senders

ChatGPT
Lists email sender identities available to the authenticated user. Use before sending email through MCP so the user can pick a permitted sender. Each entry carries replies_tracked: true means the mailbox is integrated and the recipient's reply comes back into the system; false means the address can send but nothing reads its inbox, so any reply is lost. Prefer a replies_tracked sender, and if only untracked ones exist say so before sending rather than picking one silently. It is read-only.

list_bank_statement_periods

ChatGPT
Lists periods/months for a bank statement visible to the authenticated user. Use when the user asks for monthly revenue, balances, NSF counts, DSR, fraud signals, or period-level statement details. It is read-only.

list_blasts

ChatGPT
For non-merchant Specter broker/operator users only. Lists past outreach BLASTS — SMS and email merged into one timeline, newest first. Use for "what did we send", "how did last week's blast do", or to find the blast id another blast tool needs. This is sent history, not an audience: preview_blast_audience answers who a FUTURE send would reach. Read-only.

list_capital_sources

ChatGPT
For non-merchant Specter broker/operator users only. Merchant applicant accounts must not use this tool. Lists lender-board/capital-source records visible to the authenticated user. Use before creating a lender when the user asks to add one from chat, a file, or a product guide, so you can check for an existing record by exact or near name. It is read-only and backend tenant scope/RBAC remain authoritative.

list_channel_messages

ChatGPT
Lists messages in a communication channel visible to the authenticated user. Use to retrieve SMS, email, internal notes, or thread context before answering or sending. Returns the whole channel; the endpoint has no thread filter, so pick the replies out yourself, or use list_email_records with thread_id for one email thread. Do NOT reach for list_thread_messages — that reads INTERNAL staff threads at a different route, in a different id space. It is read-only.

list_communication_channels

ChatGPT
Lists communication channels visible to the authenticated user. Each row carries TWO separate snapshots. last_message_ (at, preview, medium, direction) is the newest message in the channel WHICHEVER WAY IT WENT — it answers "what was the last thing that happened here", and when direction is inbound the preview is the customer's words, not ours. last_outbound_ (at, preview, medium) is the newest message WE SENT, kept separately so a reply since does not bury it — that is what answers "what did we last SEND them" and "have we followed up". A NULL last_outbound_at means we have never sent on this channel; say that rather than reporting their inbound message back as ours. It is the newest outbound message ROW, written when the send is QUEUED, so a send the carrier later rejected still moves it — read the message back with list_channel_messages (status, dispatch_error) before telling anyone it reached them. Both come off this ONE call across many records, so answer either question from this list instead of calling get_communication_channel_for_entity and then list_channel_messages per record. To answer "which of these records had ANY activity since X" in one call, pass workspace + entity_ids (the record IDs) + last_message_after (X); for "which of them did WE write to since X" use last_outbound_after instead — last_message_after also passes channels where only the customer wrote. Only drill into a specific channel's messages when you need more than the latest one in each direction. Read-only.

list_conversation_messages

ChatGPT
Lists messages in a conversation visible to the authenticated user. Use when a user asks for a specific SMS/email conversation thread. It is read-only.

list_email_blasts

ChatGPT
For non-merchant Specter broker/operator users only. Lists email blasts only, filterable by status and source. Email blasts are recorded automatically from bulk email sends and are read-only — there is no email equivalent of the SMS retarget tools. Prefer list_blasts unless the user asked specifically about email.

list_email_records

ChatGPT
Lists email audit records visible to the authenticated user. EXACTLY ONE entity filter applies: the handler checks lead_id, then counterparty_id, then facility_id, then submission_id, then thread_id, and stops at the first one sent — the rest are ignored SILENTLY, so send one. There is no direction, sender, recipient or limit filter and no pagination. Records come back newest first, EXCEPT when you send thread_id, which returns that thread oldest-first so it reads as a conversation. Only about the first 20 survive into what you receive, so a truncated result is never proof a record does not exist. Use when the user asks for recent email activity or lender email history. It is read-only.

list_facilities

ChatGPT
For non-merchant Specter broker/operator users only. Lists deals (facilities) across the whole book, filtered by stage, status, group, assignee, counterparty or free text. This is the tenant-wide deal list — list_my_recent_applications is only the caller's recent ones, and get_pipeline_snapshot gives counts rather than rows. To answer "how many", read pagination.total rather than counting the rows you got back: only about the first 20 survive into what you receive, so narrow with a filter rather than paging through everything. Read-only.

list_fos_setup_deal_statuses

ChatGPT
Tenant-admin-only setup helper. Lists deal/facility status definitions for the authenticated tenant before creating or changing deal settings. Leads and Merchants do not currently have tenant-editable status registries.

list_fos_setup_groups

ChatGPT
Tenant-admin-only setup helper. Lists existing tenant groups for Deals, Leads, or Merchants before creating or changing settings. Use entity_type=deals, leads, or merchants. Do not use inside Workflow Builder automation design; Workflow Builder should use read-only workflow lookup tools.

list_heartbeat_actions

ChatGPT
For non-merchant Specter broker/operator users only. Lists Heartbeat actions — proactive next steps the system proposed for leads, merchants, and deals (send a message, update a field, follow up). Defaults to status=pending (awaiting the user); use it for a daily digest or "what needs me". This operation only reads stored proposal records; it never executes, registers, discovers, or enables model-callable operations. Approval, application and reversal are separate tools already exposed in tools/list.

list_leads

ChatGPT
For non-merchant Specter broker/operator users only. Lists and filters the lead pipeline — by status, owner, tag or free-text search — for questions like which leads to work next, or how many sit at a status. Use search_records instead when the user names one specific person or business. It is read-only.

list_message_drafts

ChatGPT
Lists reply drafts awaiting human approval, optionally filtered by status or conversation. Use to check whether a conversation already has a pending draft before writing another. It is read-only.

list_my_calendar_events

ChatGPT
For non-merchant Specter broker/operator users only. The signed-in user's OWN calendar — Google or Outlook, whichever their calendar preference resolves to — for a day or a window. Answers "what is on my calendar tomorrow?", "am I free Thursday at 2?", "what meetings need me this week?". Prefer day (YYYY-MM-DD, or simply today / tomorrow / yesterday / a weekday name, resolved in the user's own time zone; days=N for a span) over from/to RFC3339 instants. Entries this CRM booked carry record {entity_type, entity_id} so you can name the lead or deal. When connected is false, repeat the answer's reason to the user verbatim — it names the setting that fixes it — and do not call anything else. Read-only.

list_my_recent_applications

ChatGPT
Lists recent financing application/facility records visible to the authenticated user. Use when the user asks what applications, deals, facilities, or funding requests they recently have in Specter. Filters are intentionally narrow: page size, recent-day window, status/stage, and search text.

list_my_tasks

ChatGPT
For non-merchant Specter broker/operator users only. Lists tasks assigned to the AUTHENTICATED user — the answer to "what is on my plate", "what do I owe today", "what am I behind on". The assignee is taken from the session and cannot be changed: to see someone else's tasks use list_tasks with assigned_to. Filter to status=pending or in_progress for open work; without a status filter this includes completed and cancelled tasks too. Read-only.

list_notes

ChatGPT
For non-merchant Specter broker/operator users only. Reads a record's notes and activity timeline, newest first — including the notes add_note writes. Use it before saying anything about what has or has not happened on a record: add_note is how notes get written, and this is the only way to read them back. To answer "how many", read pagination.total rather than counting rows. Read-only.

list_offers

ChatGPT
For non-merchant Specter broker/operator users only. Lists offers visible to the authenticated user, with their status, terms, lender, and expires_at. This is what answers "which offers are expiring", "what is outstanding on this deal", and "has anyone accepted yet" — compare_application_offers only ranks the offers on ONE facility, and there was previously no way to read offers across records at all. The endpoint has no expiry filter, no date filter, no sort and no pagination, and only about the first 20 rows survive into what you receive — the rest are replaced by a truncated marker, in no defined order. So NARROW with facility_id, submission_id, status or lender_id rather than scanning everything, then compare expires_at in what comes back; a truncated result is never evidence an offer does not exist. Read-only.

list_sms_blasts

ChatGPT
For non-merchant Specter broker/operator users only. Lists SMS blasts only, filterable by status and source. Prefer list_blasts unless the user asked specifically about texts, or you need the status/source filters this one adds. Read-only.

list_sms_records

ChatGPT
Lists SMS audit records visible to the authenticated user, filtered by facility, lead, counterparty, agent, conversation, provider, direction, or phone. There is NO submission filter and NO limit — the endpoint returns every matching record, newest first, and only about the first 20 survive into what you receive (the rest become a truncated marker), so narrow by an entity rather than scanning, and never read a truncated result as proof a record does not exist. Use when the user asks for recent SMS activity or text-message history. It is read-only.

list_submissions

ChatGPT
For non-merchant Specter broker/operator users only. Lists lender submissions across the whole book, filtered by deal, lender, status or age. This is the tenant-wide submission list; list_application_submissions covers ONE deal. Use it for "what is outstanding with lenders", "what has been sitting with a lender for a week", or "what did this lender come back on". Only about the first 20 rows survive into what you receive, so narrow with a filter rather than scanning, and never read a truncated result as proof something does not exist. Read-only.

list_system_errors

ChatGPT
System-admin only. Reads Specter platform error logs across ALL tenants for operational debugging — either the captured error records in the CRM database (source "db", the default) or live Google Cloud Logging entries from Cloud Run (source "gcp"). The backend enforces Specter system-admin access; tenant admins, brokers, operators, and merchant accounts cannot use this tool. Use it to investigate crashes, 5xx responses, panics, and recent failures: filter by level (ERROR/FATAL/WARN), a lookback window (since, e.g. "24h"), a free-text query, or a specific tenant_id. Returns each entry's time, level, message, error detail, and request id, plus (for gcp) source location, stacktrace, HTTP status, and Cloud Run revision. Read-only.

list_tasks

ChatGPT
For non-merchant Specter broker/operator users only. Lists tasks across the tenant, filtered by the record they hang off, their type, status, priority, or who they are assigned to. Use for "what is open on this deal" (entity_type + entity_id), "what is overdue", or another user's queue (assigned_to). For the caller's OWN tasks prefer list_my_tasks. To answer "how many", read pagination.total rather than counting the rows returned — that is only the page. Read-only.

list_tax_return_years

ChatGPT
Lists year-level records for a tax return visible to the authenticated user. Use when the user asks for annual tax metrics, revenue, profitability, ownership, or extracted year details. It is read-only.

list_team_members

ChatGPT
For non-merchant Specter broker/operator users only. Lists the people in this tenant — name, email, role, status — so a person named in conversation ("assign Zach", "who is Yoni") can be turned into an ID before any assignment is written. Match on email first; names repeat. Each row carries TWO different IDs and they are not interchangeable: id is the AGENT id and user_id is the USER id. The assignment tools want user_id. Read-only.

list_thread_messages

ChatGPT
For non-merchant Specter broker/operator users only. Reads the messages colleagues posted on an internal thread, newest-first. Get the thread_id from get_thread_for_entity. These are INTERNAL notes between staff — merchant-facing email and SMS live on communication channels, via list_channel_messages. Messages flagged internal-only are NOT returned here, so a thread can hold discussion this call does not show; say that rather than reporting the thread as complete. Read-only.

list_workflow_builder_catalog

ChatGPT
Reads workflow-editor reference metadata: supported triggers, node categories, action configuration, connection handles, and field schemas. These are saved workflow configuration types, not callable MCP tools. This read-only operation does not execute nodes, invoke actions, or unlock tools. All model-callable operations are already declared in tools/list.

list_workflow_facility_groups

ChatGPT
Lists the authenticated tenant's Facility/Deal groups for Workflow V3 configuration, including real group IDs, names, canonical slugs, colors, sort order, active/archive state, and default UI state. Use this before setting facility group_id or moving deals between groups/tabs. It is read-only; never guess a group_id or rely on runtime group creation when this tool can verify the existing tenant group.

list_workflow_facility_statuses

ChatGPT
Lists the authenticated tenant's Facility/Deal status registry for Workflow V3 configuration, including stable automation values, labels, colors, stages, queues, active/archive state, system flag, and terminal flag. Use this whenever a workflow condition, trigger, or update references a facility status. It is read-only; never invent status values when this tool can verify them.

list_workflow_fos_entities

ChatGPT
Lists FOS entity types available to Workflow V3 configuration for the authenticated tenant. Use before building entity-triggered workflows or field conditions. It is read-only.

list_workflow_lead_groups

ChatGPT
Lists the authenticated tenant's Lead groups for Workflow V3 configuration, including real group IDs, names, canonical slugs, colors, sort order, active/archive state, and default UI state. Use this before setting lead group_id or moving leads between groups/tabs. It is read-only; never guess a lead group_id or use a facility group ID for a lead.

list_workflow_node_types

ChatGPT
Reads the supported node configuration schemas for the tenant's workflow editor. Use to explain or configure saved workflow graphs. Node types are workflow data, not dynamically callable tools. This read-only operation never runs a node, performs its action, or changes which MCP tools are available.

list_workflows

ChatGPT
Lists the authenticated tenant's Workflow V3 workflows as a compact index: id, name, status (draft/active), trigger type, node/edge counts, and timestamps — graphs are omitted; fetch one with get_workflow. Use to find an existing workflow the user names ('publish the new-deal one', 'edit my routing workflow') before fetching, editing, or publishing it. Read-only.

onboard_lead

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL lead intake: creates the lead, creates a linked follow-up task, and optionally sends an intro SMS/email — with per-step results. PREFER THIS over separate create_lead + create_task + send_channel_message calls whenever the user asks to onboard, intake, or add-and-follow-up a new prospect. Steps run in order and stop on the first failure; the response reports exactly which steps completed. This writes live CRM/FOS data and can send an outbound message.

open_chatgpt_merchant_application

ChatGPT
Open merchant funding application in ChatGPT Opens the secure merchant funding application inside ChatGPT without a Specter login. Opening does not submit an application. The applicant enters private information and explicitly submits within the secure component, never in chat. Staff review and approval precede a separate merchant portal invitation; no workspace access is granted. GET /mcp/merchant-application-open (access: public)

operation_post_accounting_hold-clawback-rules_run_apply

ChatGPT
Apply reviewed accounting hold rules Applies hold and clawback bookkeeping changes after verifying the supplied review tokens. It does not transfer money. POST /v1/accounting/hold-clawback-rules/run (access: workspace:books)

operation_post_accounting_hold-clawback-rules_run_hold

ChatGPT
Place accounting deal hold Places a bookkeeping hold on one accounting deal with a recorded reason. POST /v1/accounting/hold-clawback-rules/run (access: workspace:books)

operation_post_accounting_hold-clawback-rules_run_preview

ChatGPT
Preview accounting hold rules Calculates proposed hold and clawback bookkeeping changes without applying them. POST /v1/accounting/hold-clawback-rules/run (access: workspace:books)

operation_post_accounting_hold-clawback-rules_run_release

ChatGPT
Release eligible accounting deal hold Releases one bookkeeping hold only when the native eligibility checks pass. POST /v1/accounting/hold-clawback-rules/run (access: workspace:books)

operation_post_accounting_hold-clawback-rules_run_save_defaults

ChatGPT
Save accounting hold rule defaults Replaces workspace hold and clawback rule defaults. Requires workspace administrator permission. POST /v1/accounting/hold-clawback-rules/run (access: workspace:books)

operation_post_boards_id_rows_act_analyze

ChatGPT
Analyze documents on selected board records Queues document analysis for files on selected records; analysis can update extracted data. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. For analysis, affected counts documents queued, not records. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_archive

ChatGPT
Archive selected board records Archives the selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_assign

ChatGPT
Assign selected board records Replaces the specified role assignment on selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_convert

ChatGPT
Convert selected board leads to deals Converts selected lead records into deals. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_copy

ChatGPT
Copy selected board records Creates copies of selected records at the specified destination. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. Skipped may also describe column values that could not be copied. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_delete

ChatGPT
Delete selected board records Deletes selected records and unlinks their linked bank connections. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_export

ChatGPT
Export selected board records Exports the selected records into a stored downloadable file. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_move

ChatGPT
Move selected board records Moves selected records to the specified destination and updates their relationships. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_status

ChatGPT
Change selected board record statuses Changes selected record statuses and records the transition history. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_boards_id_rows_act_update

ChatGPT
Update selected board record fields Replaces the specified cell values on selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/boards/:id/rows/act (access: board:read)

operation_post_definitions_key_credit-memos_id_transition_approve

ChatGPT
Approve credit memo Approves the submitted credit memo. Requires the native ledger approval permission. POST /v1/definitions/:key/credit-memos/:id/transition (access: def:read)

operation_post_definitions_key_credit-memos_id_transition_reject

ChatGPT
Reject credit memo Rejects the credit memo with a reason. Requires the native ledger approval permission. POST /v1/definitions/:key/credit-memos/:id/transition (access: def:read)

operation_post_definitions_key_credit-memos_id_transition_submit

ChatGPT
Submit credit memo for approval Submits the draft credit memo for approval. POST /v1/definitions/:key/credit-memos/:id/transition (access: def:read)

operation_post_definitions_key_rows_act_analyze

ChatGPT
Analyze documents on selected entity records Queues document analysis for files on selected records; analysis can update extracted data. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. For analysis, affected counts documents queued, not records. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_archive

ChatGPT
Archive selected entity records Archives the selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_assign

ChatGPT
Assign selected entity records Replaces the specified role assignment on selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_convert

ChatGPT
Convert selected entity leads to deals Converts selected lead records into deals. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_copy

ChatGPT
Copy selected entity records Creates copies of selected records at the specified destination. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. Skipped may also describe column values that could not be copied. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_delete

ChatGPT
Delete selected entity records Deletes selected records and unlinks their linked bank connections. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_export

ChatGPT
Export selected entity records Exports the selected records into a stored downloadable file. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_move

ChatGPT
Move selected entity records Moves selected records to the specified destination and updates their relationships. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_status

ChatGPT
Change selected entity record statuses Changes selected record statuses and records the transition history. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_act_update

ChatGPT
Update selected entity record fields Replaces the specified cell values on selected records. Select records by ids or filter, not both. Omitting both selects all visible active records. With dryRun=true, validate and preview the selection without performing the operation; false or omitted performs it. Inspect matched, affected and skipped: inaccessible or missing named records and operation-specific refusals may be skipped, so a successful response does not mean every requested record changed. POST /v1/definitions/:key/rows/act (access: def:read)

operation_post_definitions_key_rows_rowId_contract_cancel

ChatGPT
Cancel deal contract Cancels the deal contract and closes any open signature request. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_create

ChatGPT
Create deal contract record Creates the deal contract record for an accepted offer and its designated signers. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_execute

ChatGPT
Record contract execution Marks the signed deal contract as executed. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_expire

ChatGPT
Expire deal contract Expires the deal contract and closes any open signature request. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_review

ChatGPT
Mark deal contract for review Moves the deal contract record into pending review. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_send

ChatGPT
Record contract sent for signature Records that the deal contract was sent outside this operation. This records the status and recipient details; it does not send an envelope. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_sign

ChatGPT
Record received contract signature Records an existing signature from a designated signer; do not use this to sign on someone's behalf. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_contract_void

ChatGPT
Void deal contract Voids the deal contract and closes any open signature request. POST /v1/definitions/:key/rows/:rowId/contract (access: def:read)

operation_post_definitions_key_rows_rowId_funding_approve

ChatGPT
Approve funding tracking record Marks the funding tracking record approved. It does not transfer money. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_cancel

ChatGPT
Cancel funding tracking record Cancels the funding tracking record. It does not cancel a bank transfer. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_confirm

ChatGPT
Confirm funding receipt record Marks the funding tracking record confirmed. It does not transfer money. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_fail

ChatGPT
Record failed funding Marks the funding tracking record failed with the supplied reason. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_fund

ChatGPT
Record completed deal funding Records completed funding, updates deal accounting and optionally creates its repayment schedule. It does not transfer money. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_fund-and-transition

ChatGPT
Record funding and change deal status Records completed funding and changes the deal status in the same transaction, optionally creating its repayment schedule. It does not transfer money. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_process

ChatGPT
Record funding processing Records that funding is processing with an external transaction reference. It does not initiate a transfer. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_request

ChatGPT
Record funding request Opens a funding tracking record for a signed deal contract. It does not transfer money. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_definitions_key_rows_rowId_funding_reverse

ChatGPT
Record reversed funding Records a funding reversal in the deal history. It does not reverse a bank transfer. POST /v1/definitions/:key/rows/:rowId/funding (access: def:read)

operation_post_workflows_reviews_id_decide_approved

ChatGPT
Approve workflow review Approves the pending workflow review so its published run may continue with its configured actions. POST /v1/workflows/reviews/:id/decide (access: workspace)

operation_post_workflows_reviews_id_decide_rejected

ChatGPT
Reject workflow review Rejects the pending workflow review with a note and fails its waiting run. POST /v1/workflows/reviews/:id/decide (access: workspace)

patch_accounting_company-accounts_id

ChatGPT
Change a company bank account Changes purpose, packKeys, or archives (archived). A MANUAL account also takes name, institution, mask and its statement balance (statementBalanceMinor with statementOn) — the bank side of its three-way reconciliation. Accounts are archived, never deleted. Takes the admin seat (owner or admin); anyone else is refused with 403. PATCH /v1/accounting/company-accounts/:id (access: workspace:admin)

patch_accounting_deal-splits_id

ChatGPT
Change accounting deal-splits id Corrects a stake's terms. Body, all optional: basis, sharePpm, flatAmountMinor, notes. Who the stake is FOR cannot be patched — that makes it a different stake, so retire this one with post_accounting_deal-splits_id_deactivate and add its successor. Hold is its own pair of verbs so that the reason is required and the timestamps are checked in pairs. Who may: a books writer, or a team lead for a stake they can see on a deal in their reach. PATCH /v1/accounting/deal-splits/:id (access: workspace)

patch_accounting_deals_id

ChatGPT
Change accounting deals id Corrects a funded deal's agreed economics. Body, all optional: facilityId, lenderId, merchantId, reference, dealDate (YYYY-MM-DD), fundedAmountMinor, commissionMinor, psfMinor, clawbackWindowDays, slaDays, status, notes. A participant's share is NOT here — it is a stake, changed with patch_accounting_deal-splits_id or replaced by retiring the old one and adding its successor. Hold is its own pair of verbs (post_accounting_deals_id_hold, post_accounting_deals_id_release-hold) so that the reason is required. Who may: a books writer, or a team lead on a deal in their reach. PATCH /v1/accounting/deals/:id (access: workspace)

patch_accounting_receivables_id

ChatGPT
Change or cancel a receivable Changes a receivable's payer, reference, amount, date, trust account or notes, or cancels it (cancelled: true). Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. PATCH /v1/accounting/receivables/:id (access: workspace:books)

patch_blasts_id

ChatGPT
Change a blast's words Edits a blast that is still a draft: name, subject, body, goal, sendFrom, templateId, audienceFilter. An absent field leaves the column alone, and goal: "" clears the goal. sendFrom is the line the copies leave on, one of get_comms_senders; sendFrom: "" goes back to the workspace's default. goal is what a reply to the blast should lead to and is what Neo is handed when a recipient answers. subject and body may carry merge fields and either/ors ({{first_name|lead.name|"there"}}); something that looks like a field and is not one is refused here rather than sent as typed. Past draft the words are frozen with the audience — a scheduled or sent blast is the record of what a workspace said to a list of people, so the edit is refused rather than rewriting history. Who it reaches is a separate act (post_blasts_id_audience), also draft-only. PATCH /v1/blasts/:id (access: workspace)

patch_boards_id

ChatGPT
Rename a board or switch its view to table or kanban Changes a custom board's name, its purpose, the view it opens in (unchosen, table or kanban), or its recordTabs: which tabs the board's record pop-up shows besides Overview — send activity, notes, files and tasks on every board, plus calls when the rows carry a phone or email column — fewer only when the person asks for fewer. Stored deduplicated in that order; null goes back to the app's default set, and a field left out is unchanged. A board made by post_boards_import or post_boards_import_auto takes them on the import itself. Nothing about its columns or its rows moves. Needs manage on the board. purpose is the board's goal: a conversation filed on one of its records carries it as board_goal. PATCH /v1/boards/:id (access: board:manage)

patch_boards_id_actions_actionId

ChatGPT
Change a board's action button Any subset of one button's label, icon, config and enabled. Send enabled: false to take a button out of use reversibly — that is what to reach for instead of deleting it. Manage on the board. PATCH /v1/boards/:id/actions/:actionId (access: board:manage)

patch_boards_id_cells

ChatGPT
Write many rows' cells at once Merge cells into many rows of one board in a single request — the batch form of PATCH /v1/boards/:id/rows/:rowId, and what a grid paste, fill-down or cleared selection sends. Body is {rows: [{id, cells: {column name or id: text}}]}, at most 1000 rows; each row is merged, so a column a row leaves out keeps what it had. A cell is keyed by its column's name or its id (trimmed, any case) and a select cell takes its choice's value or label; a row naming a shared label, a column the board does not have or a word the column has no choice for is refused on its own line. Answers {updated, failed: [{id, error}]} — a row the rules refuse fails alone and the rest still land. A sent cell replaces what it held, so adding to Notes cells sends each row its own old text with the new — never one value over rows whose notes differ. One at a time: patch_boards_id_rows_rowId PATCH /v1/boards/:id/cells (access: board:write)

patch_boards_id_columns_columnId

ChatGPT
Rename or reconfigure a board column Any subset of one board column's label, its starred flag (whether the New record dialog asks for it up front rather than folding it under Advanced) and its per-kind config — a currency's code, a number's format, a rating's max. A column's KIND cannot be changed. Manage on the board. PATCH /v1/boards/:id/columns/:columnId (access: board:manage)

patch_boards_id_columns_columnId_options_optionId

ChatGPT
Change a dropdown choice's colour, group or label Changes one choice of a board's dropdown (select) column: its label, its colour — tone: success is green, destructive red, warning amber, secondary grey, brand and brandLight the brand colour — or the group it files under, its groupLabel band (an empty band unfiles it). The stored value never changes, so cells already carrying it stay valid. It is the call to make a choice red, or put it in a group, once an import has made it; put_boards_id_columns_columnId_options_order reorders them. Manage on the board. PATCH /v1/boards/:id/columns/:columnId/options/:optionId (access: board:manage)

patch_boards_id_rows_rowId

ChatGPT
Change one board record's cells MERGES cells into ONE row of a custom board: a column you send is overwritten, a column you leave out is left alone. A cell is keyed by its column's name or its id (trimmed, any case) and a select cell takes its choice's value or label; a label two columns share, a key no column answers to and a word the column has no choice for are each refused naming what there is, never stored as sent. The same write across many rows at once is patch_boards_id_cells; the pack-page twin is patch_definitions_key_rows_rowId. Write access on the board. A record's Notes tab draws its column labelled Notes (text or long text), so a note is a write to that cell — and a sent cell replaces what it held: to add a note, send the old text with the new under it. Many at once: patch_boards_id_cells PATCH /v1/boards/:id/rows/:rowId (access: board:write)

patch_boards_id_views_viewId

ChatGPT
Change a saved board view Any subset of one view's label, filter, sort and columnIds. Filter, sort and columnIds REPLACE what was stored rather than merging into it. The caller's own user view, or a workspace view when they manage the board. PATCH /v1/boards/:id/views/:viewId (access: board:read)

patch_calendar_events_id

ChatGPT
Change a meeting Moves or renames a meeting THIS service booked, at Google and on the record. Only the fields the body names change. An id that names no meeting of ours is a 404 - including an event sitting on the caller's calendar that was booked in Google's own UI. PATCH /v1/calendar/events/:id (access: workspace)

patch_comms_conversations_id

ChatGPT
Star, pin or archive a conversation Changes one thread's own flags — is_starred, is_pinned, is_archived — and corrects contact_name or contact_role, the name we show for the person on the other end. A field the body omits is left as it was. The ADDRESS cannot be changed: it is the thread's identity, and to reach another address you open its thread. PATCH /v1/comms/conversations/:id (access: workspace)

patch_comms_conversations_id_visibility

ChatGPT
Change who may see a conversation Sets one thread's visibility, overriding the workspace's default for that medium. The thread's owner only — anybody else who can read it is refused with 403. PATCH /v1/comms/conversations/:id/visibility (access: workspace)

patch_comms_labels_labelId

ChatGPT
Rename or recolour an inbox label Changes an inbox label's name or tone. A field the body omits is left as it was; a name the workspace already has is a 409. PATCH /v1/comms/labels/:labelId (access: workspace)

patch_comms_messages_id_status

ChatGPT
Record a message's delivery status Moves one outbound message to sent, delivered, read, failed or cancelled, stamping delivered_at and read_at the first time each is reached. For a caller holding OUR message id; a device reporting by the carrier's id uses POST /v1/comms/outbound/receipt, which writes the same columns. A word the schema does not know is refused, never quietly stored. PATCH /v1/comms/messages/:id/status (access: workspace)

patch_credit-memos_id

ChatGPT
Change credit-memos id Edits a credit memo, DRAFTS only: the store puts that in the WHERE, so a memo already submitted reads as a not-found rather than being quietly rewritten under its reviewer. Body, all optional: title, memoType, any of the nine narrative panes, requestedAmountMinor, recommendedAmountMinor (or clearRecommendedAmount: true to recommend no figure at all), recommendedRate, recommendedTermMonths, recommendedProduct, recommendedDecision, recommendedConditions, expiresAt. A memo in review is changed by rejecting it and writing its successor, not by patching. PATCH /v1/credit-memos/:id (access: workspace)

patch_definitions_key

ChatGPT
Edit record page settings Renames a record page or changes its icon, and carries the page's switches: tagsEnabled; recordsEnabled, which decides only whether the board draws the sub-record dropdown under a row — the relations themselves are untouched either way; and view (0252, 0440), table, kanban or cards, how the page opens for every seat — any other word is refused. An absent field is left alone, a blank or over-120-character name is refused, and the answer is the page as it now stands. Workspace admin; this changes the page's SHAPE, not any record on it. PATCH /v1/definitions/:key (access: def:admin)

patch_definitions_key_actions_actionId

ChatGPT
Edit a record page action button Changes one action-bar button: label (up to 60 characters), icon, config and enabled. A disabled action stays on the bar and stops being offered, which is the reversible way to take a button out of use. An absent field is left alone; answers 204. Workspace admin. PATCH /v1/definitions/:key/actions/:actionId (access: def:admin)

patch_definitions_key_cells

ChatGPT
Write many records' cells at once Merge cells into many records on one page in a single request — the batch form of PATCH /v1/definitions/:key/rows/:rowId, and what a grid paste, fill-down or cleared selection sends. Body is {rows: [{id, cells: {column name or id: text}}]}, at most 1000 rows; each row is merged, so a column a row leaves out keeps what it had, and every record still gets its own notification and activity entry. A cell is keyed by its column's name or its id (trimmed, any case) and a select or status cell takes its choice's value or label; a row naming a shared label, a column the page does not have or a word the column has no choice for is refused on its own line. Answers {updated, failed: [{id, error}]} — a record the rules refuse fails alone and the rest still land. A sent cell replaces what it held, so adding to Notes cells sends each row its own old text with the new — never one value over rows whose notes differ. One at a time: patch_definitions_key_rows_rowId PATCH /v1/definitions/:key/cells (access: def:read)

patch_definitions_key_columns_columnId

ChatGPT
Edit a record page column Changes one column's presentation: label (up to 80 characters), defaultEnabled, minWidth (40-800), alignEnd, starred (whether the Add-row dialog asks for this column up front) and config. A patch CANNOT change a column's kind — config is validated against the kind already on the row — and an absent field is left alone. Answers 204; the column id must be on this page. Workspace admin. PATCH /v1/definitions/:key/columns/:columnId (access: def:admin)

patch_definitions_key_columns_columnId_options_optionId

ChatGPT
Relabel or recolour one choice on a page Changes one select option's label, its tone, or the groupLabel band it files under — an empty band unfiles it. The stored value never changes, so cells already carrying it stay valid. Workspace admin. PATCH /v1/definitions/:key/columns/:columnId/options/:optionId (access: def:admin)

patch_definitions_key_entity-groups_groupId

ChatGPT
Rename a relationship or set its principal Renames, re-kinds or re-notes one entity group, and sets or clears its PRINCIPAL — the member everything else hangs off, which must already be in the group. An absent field is left alone; an explicit null on primaryRowId clears the principal. PATCH /v1/definitions/:key/entity-groups/:groupId (access: def:read)

patch_definitions_key_groups_groupId

ChatGPT
Rename, recolour or file a record group Renames, re-describes or recolours a record group, moves it with position, and files it under a status section with statusGroup (a name, or null to unfile it — omitting the field leaves the filing alone). At least one field must be given. Answers the group with its member count; the records in it are untouched. PATCH /v1/definitions/:key/groups/:groupId (access: def:read)

patch_definitions_key_rows_rowId

ChatGPT
Update fields on a record MERGES the given cells into one record and answers the whole row: a column the body does not mention keeps its value, so a single-cell edit cannot blank the rest of the record with a stale copy. A cell is keyed by its column's name or its id (trimmed, any case) and a select or status cell takes its choice's value or label; a label two columns share, a key no column answers to and a word the column has no choice for are each refused naming what there is, never stored as sent. Use post_definitions_key_rows_rowId_status for the status word — it has its own door because it also writes the move to the record's timeline — and patch_definitions_key_cells for the same change across many rows. Money-board rows are refused; an unknown row is a 404. A record's Notes tab draws its column labelled Notes (text or long text), so a note is a write to that cell — and a sent cell replaces what it held: to add a note, send the old text with the new under it. Many at once: patch_definitions_key_cells PATCH /v1/definitions/:key/rows/:rowId (access: def:read)

patch_definitions_key_rows_rowId_debt-schedules_scheduleId

ChatGPT
Edit a business debt schedule Rewrite one business debt schedule. Same body and the same revision check as creating it — this is the door an edit goes through, and a stale revision loses. PATCH /v1/definitions/:key/rows/:rowId/debt-schedules/:scheduleId (access: def:read)

patch_definitions_key_rows_rowId_folders_folderId

ChatGPT
Rename or move a document folder Change a folder's name, its parent, or its position. An absent field is left alone; parentFolderId sent as null moves the folder to the record's Documents root. A folder cannot be moved inside its own subtree. PATCH /v1/definitions/:key/rows/:rowId/folders/:folderId (access: def:read)

patch_definitions_key_rows_rowId_notes_noteId

ChatGPT
Edit a record note Update one internal note using its current version and a nonempty body. Preserves the original author and creation time, records the editor and edit time, and rejects stale versions. Refresh notes before retrying a conflict. Requires write access to the record. PATCH /v1/definitions/:key/rows/:rowId/notes/:noteId (access: def:read)

patch_definitions_key_scripts_scriptId

ChatGPT
Create a revised call script version Edits a call script, which in this table means writing the NEXT VERSION: the live version is retired and a new one written in the same transaction, so the ladder never has two live rungs and never has none. It answers 201 with the new script, because the id the caller sent is not the id it now holds. A field not sent is inherited. The id may name any version — the ladder continues from the LIVE one. Leads page, admin. PATCH /v1/definitions/:key/scripts/:scriptId (access: def:admin)

patch_definitions_key_statuses_statusId

ChatGPT
Change a status's colour, group or tab, or archive it Edits one status of a page's status column (Leads, Deals): label, its colour — tone: success is green, destructive red, warning amber, secondary grey, brand and brandLight the brand colour —, groupLabel (the tab it files under), isTerminal, groupId (a record group — a section — on this page, or null to file it under nothing) and isActive. It is the call to make a status red, or put statuses in a group or under a tab. isActive: false ARCHIVES the status — records already on it keep it and still read back, but a move onto it is refused — which is the reversible alternative to deleting it; true restores it. An absent field is left alone and the answer is 204. Workspace admin. PATCH /v1/definitions/:key/statuses/:statusId (access: def:admin)

patch_definitions_key_threads_threadId

ChatGPT
Edit a discussion thread subject Changes a thread's subject, and nothing else. Resolving is deliberately not patchable here — it would be a second, unlabelled way to settle a thread that skips the timeline entry the resolve route writes. Answers the thread. PATCH /v1/definitions/:key/threads/:threadId (access: def:read)

patch_definitions_key_threads_threadId_messages_messageId

ChatGPT
Edit your discussion message Rewrites one message. THE AUTHOR ONLY — anyone else's message is a 404 — and the row keeps an edited stamp so the app can say it was changed. A blank body is refused. PATCH /v1/definitions/:key/threads/:threadId/messages/:messageId (access: def:read)

patch_definitions_key_underwriting_id

ChatGPT
Update an underwriting record Writes the fields the underwriting tab's inline editors change on one underwriting record: engagementStatus and creditWorthinessDecision, the money figures in minor units (eligible loan amount, monthly payment, guaranty fee, closing costs, working capital, debt refinance, annual revenue, existing debt, personal revolving balance and limit), the rates, the scores (SBSS, FICO, time in business, employees), the milestone dates, and the underwriter and closer. :id names the UNDERWRITING record, not the deal — its id already names one deal. A patch naming nothing is refused rather than treated as a read. Deals page only. PATCH /v1/definitions/:key/underwriting/:id (access: def:read)

patch_definitions_key_views_viewId

ChatGPT
Edit a saved record page view Edits a saved view's label, filter, sort or columnIds; an absent field is left alone. A user view is editable only by the person who owns it and a workspace view only by a workspace admin — anyone else's view is a 404, never a 403. Answers 204. PATCH /v1/definitions/:key/views/:viewId (access: def:read)

patch_files_id

ChatGPT
Rename or classify a file Rename a file (name, trimmed, 1–255 characters, taken as given — keep the extension yourself) and/or say what a document IS: documentType from the closed vocabulary (bank_statement, tax_return, credit_report, contract, … other), the span it covers as YYYY-MM-DD periodStart/periodEnd, and statementStatus (new, submitted or fraud) on a bank statement. A body with only name renames and leaves the classification alone; any body without name is a classification, and each of its fields is read as given rather than merged, so an empty string puts the file back to unclassified. A word outside the vocabulary is refused rather than stored. Metadata only — nothing is sent to anybody and no storage is needed. Answers the file with its links and the caller's star. PATCH /v1/files/:id (access: workspace)

patch_forms_id

ChatGPT
Change a form Any subset of a form's name, description, fields, sections, settings and targetDefinitionKey. Fields, sections and settings REPLACE, whole — send the complete list, not just the line you changed. The SLUG is deliberately not patchable: it is the published address, and changing it would break every link already handed out. PATCH /v1/forms/:id (access: workspace)

patch_me_calendar

ChatGPT
Change your calendar settings Changes some of the caller's calendar preferences - calendarId, workDayStartMinute, workDayEndMinute, defaultDurationMinutes, mirrorTaskDueTimes, onTaskCompleted (cancel | mark_done | keep) and reminderMinutes (0, 10, 30 or 60). A field the body omits is left as it was. PATCH /v1/me/calendar (access: workspace)

patch_message-automation-rules_id

ChatGPT
Change a standing message rule Edits one standing rule: name, description, enabled, triggerEvent, entityKind, triggerCondition, templateId (or clearTemplate: true to leave it naming no template), requiresApproval, recipientField. An absent field leaves the column alone. enabled: false is how a rule is switched off and kept — the reversible alternative to delete_message-automation-rules_id. PATCH /v1/message-automation-rules/:id (access: workspace)

patch_message-drafts_id

ChatGPT
Change message-drafts id Edit a saved draft. Legacy partial edits take recipient, subject or body and only accept draft status. Inbox email edits provide expectedUpdatedAt (the row's current updatedAt), recipient, subject, body and accountIdentifier (empty to choose a mailbox later). Versioned edits are restricted to the caller's own email drafts, reject stale revisions, and reopen an approved-but-unsent draft for review. Queued or cancelled drafts cannot be edited. The chosen mailbox must be in get_comms_senders. Saving sends nothing. PATCH /v1/message-drafts/:id (access: workspace)

patch_message-templates_id

ChatGPT
Change a message template Edits one template's words: name, subjectTemplate, greetingTemplate, bodyTemplate, closingTemplate, bodyFormat, entityKind, isActive, isDefault. An absent field leaves the column alone. Channel and category cannot be patched — a template that belongs to the other channel is a new one. isActive: false archives it: it stays readable and leaves the picker, which is the gentle alternative to deleting it. Editing a template never changes words already sent, because a draft and a blast each keep their own frozen copy. PATCH /v1/message-templates/:id (access: workspace)

patch_neo_memories_id

ChatGPT
Change something Neo remembers Changes one of the caller's memories by its id (read the list first): its text, its kind (rule or fact), or both — "actually, send from Josh instead", "that is not a rule, just a fact". A field the body leaves out stays as it was. Another person's id is a 404. PATCH /v1/neo/memories/:id (access: workspace)

patch_notification-preferences

ChatGPT
Change your notification settings Turns notification types on or off for the caller, in-app and sound, one entry per type in {updates: [{notification_type, ...}]}. A type the body does not name is left as it was. PATCH /v1/notification-preferences (access: workspace)

patch_obligations_id

ChatGPT
Change an obligation Edits one obligation in place — retitle it, reword it, move its due date, hand it to somebody else, change its priority or its amount. Body, every field optional: title, description, notes, priority, dueAt, assignedToUserId, amountMinor, slaHours. An absent field leaves the column alone; emptying a nullable one takes its own flag — clearAssignee, clearAmount, clearSla. Moving an obligation's STATUS is not done here: complete, waive and mark-overdue are their own routes because each stamps its own evidence. PATCH /v1/obligations/:id (access: workspace)

patch_reports_views_id

ChatGPT
Change a saved report Replaces a saved report's question and its wording in ONE body — send name, description, visibility and definition together, not a field at a time, because half a definition leaves a report nobody can reason about. The report's owner or an admin; a built-in is nobody's and is duplicated rather than edited, and somebody else's private report answers 404 rather than 403. PATCH /v1/reports/views/:id (access: workspace)

patch_tasks_id

ChatGPT
Change a task Edits one task: title, description, status, priority, assignee (assigneeAccountId from the roster, or assigneeUserId) and dueAt. An absent field leaves the column, and emptying a nullable one has its own flag — clearAssignee, clearDueAt. cancellationReason carries the sentence that rides with a cancel. Re-dating a task MOVES the meeting it mirrors rather than booking a second one. PATCH /v1/tasks/:id (access: workspace)

patch_workflows_id

ChatGPT
Edit a workflow draft Edits a workflow: name, description, definition or status, each on its own, so renaming one does not mean sending the canvas back. Editing is SAFE and takes effect on nothing — a run executes the published version, never the draft — until somebody publishes. This is also how a workflow that HAS published is retired: {"status": "archived"}, because delete_workflows_id only takes a draft that never published. PATCH /v1/workflows/:id (access: workspace:admin)

patch_workspace

ChatGPT
Change workspace Rename the workspace. Admin only — a name is shape rather than data — and 403 when the caller's role cannot. Nothing else about the workspace is settable here: the tier is billing's and the company's paperwork is KYC's. PATCH /v1/workspace (access: workspace:admin)

patch_workspace_compliance_reports_id

ChatGPT
Change a compliance report Admin only, and a whole replacement rather than a merge. Editing the question does NOT change what past runs claim to have examined: every run carries the checks it actually ran. Setting active to false parks the report without losing its evidence. PATCH /v1/workspace/compliance/reports/:id (access: workspace:admin)

patch_workspace_inbound_id

ChatGPT
Re-aim or pause an inbound address Changes where an inbound address routes, or switches it off without deleting it. The address itself never changes - a different address is a different place. PATCH /v1/workspace/inbound/:id (access: workspace:admin)

patch_workspace_members_accountId

ChatGPT
Change workspace members accountId Change one member's seat: owner, admin, team_lead or member. ADMIN ONLY, and an owner seat is an owner's to give — anyone lesser trying is a 403. 409 when the change would leave the workspace with no owner at all. Taking the seat away entirely is the DELETE on the same address. PATCH /v1/workspace/members/:accountId (access: workspace:admin)

patch_workspace_role-presets_presetId

ChatGPT
Edit a role preset Changes a preset's name, description, seat, pages (board template -> none | read | write) or accounting (read | manage | ""). An outside affiliate's preset takes a name and description only — what an outsider is shown is the pack's. Editing changes nobody until the preset is applied again. A workspace admin's. PATCH /v1/workspace/role-presets/:presetId (access: workspace:admin)

patch_workspace_sidebar-folders_id

ChatGPT
Rename or move a sidebar folder Renames a folder (label, refused as a conflict when another folder already has that name) and/or moves it to index position among the folders, the rest closing up around it. An absent field is left alone; a rename is recorded on the workspace's activity feed. Any seat. PATCH /v1/workspace/sidebar-folders/:id (access: workspace)

patch_workspace_sla-policies_id

ChatGPT
Change a stage's SLA Admin only. A whole replacement, not a merge: send what the read answered with your changes folded in, because merging would make "clear the working days" and "leave the working days alone" the same request. Setting active to false parks the policy, which then measures nothing and rings nothing without losing what was promised. PATCH /v1/workspace/sla-policies/:id (access: workspace:admin)

patch_workspace_teams_teamId

ChatGPT
Change workspace teams teamId Rename a team, change its lead, or re-parent it — all three optional. The EMPTY STRING is how a field is cleared: "" for leadAccountId leaves the team leaderless, "" for parentTeamId puts it back at the top. Admin only; 409 on a duplicate name. PATCH /v1/workspace/teams/:teamId (access: workspace:admin)

patch_workspace_underwriting_rule-sets_setId

ChatGPT
Change an underwriting rule set Change a rule set's name, description, or which stage and product it applies to; or retire it with status "archived". It never republishes - what is running keeps running until the publish route is called. There is no delete: an evaluation names a version of this set and its reasoning must stay readable. PATCH /v1/workspace/underwriting/rule-sets/:setId (access: workspace:admin)

patch_workspace_underwriting_rule-sets_setId_rules_ruleId

ChatGPT
Change an underwriting rule Change one draft rule - its condition, outcome, reason, weight, order, or whether it is active. The published version is untouched, which is what makes editing a live set safe. PATCH /v1/workspace/underwriting/rule-sets/:setId/rules/:ruleId (access: workspace:admin)

patch_workspace_vendors_id

ChatGPT
Change a vendor Admin only, and a whole replacement rather than a merge. Setting active to false RETIRES the vendor: it keeps the deals that named it readable, stops it filling any new slot, and takes it out of the contract-currency check. PATCH /v1/workspace/vendors/:id (access: workspace:admin)

post_accounting_allocations

ChatGPT
Allocate a bank transaction Records that this much of that bank line was this deal's money — the write that moves the deal's collected, outstanding and every participant's available figure, at read time. Body: dealId (required), amountMinor, allocationType (one of commission, psf, clawback_recovery, bonus, management_fee, syndication_fee), and optionally bankTransactionId, allocatedOn (YYYY-MM-DD) and notes. Over-allocating a bank line or a deal is refused with 409 by a guard inside the INSERT, and there is no override parameter — that guard is the one rule stopping a deposit being counted twice. Money LEAVING is not an allocation: a draw against a stake is post_accounting_payouts. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. POST /v1/accounting/allocations (access: workspace:books)

post_accounting_bank-transactions

ChatGPT
Record a bank transaction Imports a batch of bank lines, which is how money that has actually arrived enters the books. Body: transactions, an array of at least one and at most 1000 entries of {externalId, source, postedOn (required, YYYY-MM-DD), currency, amountMinor, accountName, accountLast4, description, memo}. The whole batch runs in one transaction and it is idempotent by externalId — a line already present is skipped, not refused, so importing August twice leaves August alone; the answer says how many were imported and how many skipped. Importing a line credits no deal on its own: that is a separate allocation (post_accounting_allocations). Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A hand import belongs to no linked account, so it needs a grant that reaches every account. POST /v1/accounting/bank-transactions (access: workspace:books)

post_accounting_categories

ChatGPT
Create a transaction category Add a NEW workspace category, placed at the end of the workspace's order. The identifier is lowercase letters, digits and underscores, starting with a letter, and must not already exist — an existing identifier is refused (409) rather than overwritten; to change a category, save it by its identifier instead. Kind is one of expense, income, transfer, asset, liability or equity. Keywords are vendor or memo words that suggest this category for a bank line. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. POST /v1/accounting/categories (access: workspace:books)

post_accounting_company-accounts

ChatGPT
Add a manual bank account Registers a bank Plaid cannot reach. Body: name, purpose (operating | trust | escrow), optional institution, mask (last four), currency, packKeys (only a trust or escrow account is tagged). Takes the admin seat (owner or admin); anyone else is refused with 403. POST /v1/accounting/company-accounts (access: workspace:admin)

post_accounting_deal-costs

ChatGPT
Record a deal cost Records a cost against a funded deal, reducing its net revenue. Body: dealId (required), category, description, vendor, currency, amountMinor. There is no chart of accounts in this workspace: spend is a typed row against the deal whose revenue it reduces, not a bucket on a chart. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. POST /v1/accounting/deal-costs (access: workspace:books)

post_accounting_deal-costs_id_paid

ChatGPT
Mark a deal cost paid Marks a cost paid and stamps the moment. No body. Every net-basis stake on that deal moves with it, because a cost reduces net revenue and the availability view reads costs live. To take a cost out of the books instead of settling it, post_accounting_deal-costs_id_void. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. POST /v1/accounting/deal-costs/:id/paid (access: workspace:books)

post_accounting_deal-splits

ChatGPT
Create a deal split Lays down a stake: who participates in a funded deal's money, on what basis, for how much. Body: dealId (required) and who it is for — accountId, crmUserId or externalParty — plus partyKind, basis, and exactly ONE measure, either sharePpm (parts per million, so 300000 is 30%) or flatAmountMinor; notes is optional. A share is measured against the commission, the PSF, gross, the funded amount, or net (gross less every commission-basis entitlement, less costs, less any open clawback). A syndicator is a stake with partyKind: syndicator and basis: funded. To REPLACE a stake, retire the old one first with post_accounting_deal-splits_id_deactivate — the unique index refuses two live stakes for one party. Who may: a books writer for any party; a team lead only on a deal in their reach and only for accountId naming themselves or someone they lead — a referral partner's, the house's or another team's stake is the books' to lay. POST /v1/accounting/deal-splits (access: workspace)

post_accounting_deal-splits_id_deactivate

ChatGPT
Deactivate a deal split Retires a stake: its status becomes inactive and the moment is stamped, so the record of who was on the deal survives its replacement and the unique index is free for a successor. There is no DELETE on a stake, deliberately — this is the route to call before adding the stake that replaces it. Who may: a books writer, or a team lead for a stake they can see on a deal in their reach. POST /v1/accounting/deal-splits/:id/deactivate (access: workspace)

post_accounting_deals

ChatGPT
Add a deal to the ledger Records the agreed economics of a funded deal — the row every "collected", "outstanding" and "available" figure is then computed from. Body: dealDate (required, YYYY-MM-DD) and the deal it mirrors, which must be facilityId or dealRowId (a deal record on our own board) or the request is refused; then lenderId, merchantId, reference, currency, fundedAmountMinor, commissionMinor, psfMinor, clawbackWindowDays, slaDays, notes. Money is integer minor units. Only what was AGREED is stored here; who is owed a share of it is a separate call to post_accounting_deal-splits, and commission with no stake behind it surfaces as a no_splits exception. Who may: a books writer for any deal; a team lead only with dealRowId naming a deal record they or their team are on. A plain member is refused. POST /v1/accounting/deals (access: workspace)

post_accounting_invoices

ChatGPT
Draft an invoice from a record Opens a draft invoice on one record (rowId, any page's row) — use it for "invoice Maria $150 for a cleaning", "bill this client", "send them a payment request". Named by a customer rather than a record: find the record first with get.search (the customer's name) and pass its row id. Optional: lines (each description, quantityMilli in thousandths — 1000 is one, 1500 is 1.5 — unitMinor in cents, so $150 is 15000, taxable), taxPpm (82500 is 8.25%), dueOn (YYYY-MM-DD), billToName, billToEmail, memo, paymentInstructions, currency. Leave out what was not said: the due day (today plus the workspace's terms), the tax and how to pay come from Accounting › Settings › Invoicing (get.accounting.invoice-settings), who it bills from the record's person and client email; get.accounting.invoice-lines.recent has the prices this workspace usually charges. A draft has no number and no link; issue it with post.accounting.invoices.id.issue, then send it. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. POST /v1/accounting/invoices (access: workspace:books)

post_accounting_invoices_id_issue

ChatGPT
Issue an invoice Freezes a draft into what the client is sent, in one step: the next number from the workspace's counter, the paper rendered through the workspace's invoice template onto the record, its PDF filed on the record, the client's link minted, the receivable booked for the total, and invoice.issued on the record's timeline. Body (optional): issuedOn (default today), dueOn (default the draft's; one of the two must say it). Refused with 400 when there are no lines, the total is zero or nobody is billed, and with 409 when it was already issued. Issuing sends nothing: send it with post.accounting.invoices.id.send (email) — "issue and send" is both calls. Takes the books to KEEP; anyone else is refused with 403. POST /v1/accounting/invoices/:id/issue (access: workspace:books)

post_accounting_invoices_id_payments

ChatGPT
Record a payment on an invoice Books a payment the firm received against an issued invoice — "mark INV-1003 paid by check", "they paid $200 cash": find the invoice by its number with get.accounting.invoices ?q=, then method (check | ach | wire | card | cash | transfer | online | other), optional amountMinor (default the whole balance), paidOn (default today), reference, memo. More than the balance is refused with 400. The payment that settles the invoice writes invoice.paid on the record — the event a rule fires on. Records what happened at the bank; it never moves money. Takes the books to KEEP; anyone else is refused with 403. POST /v1/accounting/invoices/:id/payments (access: workspace:books)

post_accounting_invoices_id_payments_entryId_reverse

ChatGPT
Put back an invoice payment Reverses one recorded payment — a bounced check, a payment recorded by mistake, a refund — as a new journal entry naming it, so the invoice owes that much again. entryId is the payment's entry in get.accounting.invoices.id's entries. Body (optional): memo. Takes the books to KEEP; anyone else is refused with 403. POST /v1/accounting/invoices/:id/payments/:entryId/reverse (access: workspace:books)

post_accounting_matters_rowId_settlement_close

ChatGPT
Close a matter from its settlement Marks the case closed when every line of the approved statement is paid and the matter's trust ledger is zero, moving the row to status (one of its page's closing statuses) through the row's own status door, in one transaction. Refused with 409 naming the first check that fails. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/matters/:rowId/settlement/close (access: workspace:books)

post_accounting_matters_rowId_settlement_draft

ChatGPT
Start a settlement statement Opens the matter's next statement version as a draft: a copy of the latest version, or a first draft from the row's own settlement amount, fee share and linked providers. One draft at a time; asking again answers the open draft. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/matters/:rowId/settlement/draft (access: workspace:books)

post_accounting_matters_rowId_trust_deposits

ChatGPT
Record a deposit into trust Records money received into a trust or escrow account for a matter. Body: trustAccountId, amountMinor, entryOn, party (who paid), method (check | wire | ach | transfer | cash | other), optional reference, memo, receivableId. The account must serve the matter's pack. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/matters/:rowId/trust/deposits (access: workspace:books)

post_accounting_me_payout-methods_id_account-number_reveal

ChatGPT
Reveal an authorized account number Explicitly reveal the stored account number for this authorized record. Returns available false with a reason when the original number was not retained; never reconstructs a number from its mask. Payout methods are restricted to their owner. POST /v1/accounting/me/payout-methods/:id/account-number/reveal (access: workspace)

post_accounting_me_payout-methods_id_default

ChatGPT
Make a payout method the default Makes one of the CALLER's own payout methods the one their next payout request names. No body. Somebody else's method is a not-found. POST /v1/accounting/me/payout-methods/:id/default (access: workspace)

post_accounting_payouts

ChatGPT
Request a payout Asks to draw money against ONE stake — the request, not the payment. Body: splitId (required; get_accounting_me_availability names the caller's own stakes), amountMinor (more than zero) and notes. The available balance is read inside the INSERT and snapshotted onto the row, so the guard and the audit figure cannot disagree and two concurrent requests for one balance cannot both pass. Asking for more than is available reads as a not-found on the stake, deliberately: the alternative tells a caller how much somebody else's stake holds. The stake must be the CALLER's own. The one exception: a books writer may raise a request on a stake that names no seat (a referral partner, a syndicator, the house); never on a colleague's. The request records the caller's default payout method at that moment (get_accounting_me_payout-methods). POST /v1/accounting/payouts (access: workspace)

post_accounting_payouts_id_approve

ChatGPT
Approve a payout Approves a payout request: the status becomes approved and the decider and the moment are stamped. The admin seat. Nobody may decide their own request — that is in the UPDATE's WHERE, so approving your own reads as a not-found. Approving does not move any money; recording that it moved is post_accounting_payouts_id_paid. POST /v1/accounting/payouts/:id/approve (access: workspace:admin)

post_accounting_payouts_id_cancel

ChatGPT
Cancel a payout request Cancels a payout request. The requester's own act, so it takes the ordinary write door rather than the admin one. No body. A request that was approved and then failed to settle is cancelled here, because there is no failed state on a payout — what failed to settle is still owed. POST /v1/accounting/payouts/:id/cancel (access: workspace)

post_accounting_payouts_id_paid

ChatGPT
Mark a payout paid Records that the money for an approved payout actually left, stamping the moment and, optionally, the bank line it went out on. The admin seat. Body: bankTransactionId, optional. This states a fact about money that has already moved — sending it is not this service's job, so there is no disbursement here. There is no failed state on a request either: a settlement that failed leaves the request owed, so cancel it or leave it in the queue. POST /v1/accounting/payouts/:id/paid (access: workspace:admin)

post_accounting_payouts_id_reject

ChatGPT
Reject a payout Turns a payout request down. The admin seat. Body: reason, required — a rejection needs a reason the person can read. Nobody may decide their own request, so rejecting your own reads as a not-found. Not the admin's own request — withdraw that with post_accounting_payouts_id_cancel. POST /v1/accounting/payouts/:id/reject (access: workspace:admin)

post_accounting_plaid-connections_itemId_check-payment

ChatGPT
Check an expected bank payment Compare posted transactions against dynamic expectedAmountMinor, dueAt (RFC3339), graceHours, currency, query and direction. Returns matched, not_due, not_observed or unknown. A not_observed result is a review signal, not proof of nonpayment. Unknown or stale data must not trigger a demand. POST /v1/accounting/plaid-connections/:itemId/check-payment (access: workspace:admin)

post_accounting_plaid-connections_itemId_refresh

ChatGPT
Request fresh bank transactions Request an institution refresh for a connected bank. Returns 202 refreshRequested, not completed data. Changes arrive through the Plaid webhook and sync. Five-minute per-bank cooldown returns 429. Requires Plaid Transactions Refresh enabled. POST /v1/accounting/plaid-connections/:itemId/refresh (access: workspace:admin)

post_accounting_plaid-connections_itemId_sync

ChatGPT
Sync a company bank feed Queues a durable sync of this company bank and answers queued=true. The worker imports posted transactions from selected accounts and resumes pagination automatically. Use bank_link.bank_synced to continue after completion; do not interpret the queued response as imported data. POST /v1/accounting/plaid-connections/:itemId/sync (access: workspace:admin)

post_accounting_plaid-connections_itemId_transactions_search

ChatGPT
Search imported bank transactions Read bank transactions with currency, date range or lookbackDays, accountId, query and credit/debit direction. Signed minor units: credits positive. Matched count and total cover all matching rows; transactions are capped at 500 and truncated is explicit. lastSyncedAt names the last complete import, not a guarantee the institution is current. POST /v1/accounting/plaid-connections/:itemId/transactions/search (access: workspace:admin)

post_accounting_receivables

ChatGPT
Add a receivable Writes down money a matter expects. Body: rowId (any page's row), payer, expectedMinor, optional kind (settlement | fee | costs | other), reference, expectedOn, trustAccountId (a trust account that serves the row's pack), notes. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/receivables (access: workspace:books)

post_accounting_settlement_lines_lineId_disburse

ChatGPT
Record a payment from trust Records that one line of an approved statement was paid from its trust account. Body: entryOn, method, optional amountMinor (default: what is left on the line), reference (check or wire number), memo. The fee and advanced expenses are paid to the firm's operating account. Refused with 409 until the client has signed the approved statement, past the line's payable, or below a zero ledger. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/settlement/lines/:lineId/disburse (access: workspace:books)

post_accounting_settlement_statements_id_approve

ChatGPT
Approve a settlement statement Freezes a draft as the statement trust pays from. Refused unless every line sums to the gross to the cent, the client's net is not negative, every required check but the client's is ticked and a trust account is named. The client approves AFTER this, by signing the approved statement (post.accounting.settlement.statements.id.document), and trust pays no line until they have. Takes the admin seat (owner or admin); anyone else is refused with 403. POST /v1/accounting/settlement/statements/:id/approve (access: workspace:admin)

post_accounting_settlement_statements_id_checks_checkId

ChatGPT
Tick a readiness check Ticks or unticks one box on a draft's readiness checklist. Body: confirmed, optional note. Who and when are recorded. settlement_received ticks only when the matter's trust deposits cover the gross. client_approved is refused with 400: the client's own signature on the approved statement ticks it. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/settlement/statements/:id/checks/:checkId (access: workspace:books)

post_accounting_settlement_statements_id_document

ChatGPT
Issue a closing statement for the client to sign Renders the matter's APPROVED settlement statement through the workspace's settlement_statement template and stores it as a paper on the matter, keeping the statement's id in its substitutions. Send it with post.documents.id.signing; when that envelope completes, the statement's client_approved check is ticked by the signature, and only then does trust pay any line. A draft, superseded or closed statement is refused with 409. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. POST /v1/accounting/settlement/statements/:id/document (access: workspace:books)

post_accounting_trust_entries_id_match

ChatGPT
Match a trust entry to its bank line Reconciles one journal entry to the bank line that proves it. Body: bankTransactionId. The line must be on the trust account's own linked bank account, the same direction and amount, not flagged for review, and matched to no other entry. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/trust/entries/:id/match (access: workspace:books)

post_accounting_trust_entries_id_reverse

ChatGPT
Reverse a trust entry Records the correction of an entry: the same amount the other way, naming what it cancels. Body: entryOn, optional memo. Refused when it would take the matter's trust ledger below zero. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. Records what the firm did at its bank; it never moves money. POST /v1/accounting/trust/entries/:id/reverse (access: workspace:books)

post_billing_cancel

ChatGPT
Cancel at the end of the period Stops the subscription renewing and answers the status and the date it ends. Nothing is cancelled immediately and the plan does not change: the workspace paid for this period and keeps every entitlement until it runs out. Needs the workspace's owner or admin seat. POST /v1/billing/cancel (access: workspace:admin)

post_billing_portal-session

ChatGPT
Open the Stripe customer portal Answers a one-time url for Stripe's own hosted portal, where a person changes the card, reads invoices and cancels. We deliberately do not rebuild any of that. A workspace with no Stripe customer gets one made rather than a refusal. Needs the workspace's owner or admin seat. POST /v1/billing/portal-session (access: workspace:admin)

post_blasts

ChatGPT
Compose a blast to many people Composes a blast: ONE set of words aimed at many people, the same thing said to everyone on a list — this is the route for "email this whole list the same thing" or "text all of these people". Body: kind ("sms" or "email"), name (required, one per kind), body (the words), subject (email only, and required there), goal, sendFrom, templateId, audienceFilter (the rule that chose the list, kept for the reader and never re-run) and retargetOfBlastId. goal is what a reply should lead to, in a sentence — "book a 10-minute call about renewing; never quote a rate". Neo is handed it with every thread a copy is sent in, so it decides how replies are answered: always set one, and ask the person for it if they have not said. The words may carry merge fields, filled per recipient: {{first_name}} is a field and {{first_name|lead.name|"there"}} is an either/or — each step is tried in turn and the quoted one is used when none had a value. A recipient with nothing to fill a field and no quoted fallback is SKIPPED rather than sent a gap, so give every field a fallback (get_blasts_merge-fields lists the names). sendFrom is the line its copies leave on — a phone number for a text blast, a mailbox for an email one, taken from get_comms_senders; leave it out for the workspace's default line, and name it whenever the person says which number or mailbox to use. A line the caller may not send from is refused. It starts as a draft with NO audience and no time: aim it with post_blasts_id_audience_board, read get_blasts_id_preview, and post_blasts_id_dispatch is the verb that actually sends it. For one message to one person, post_comms_send. Messages worded for each person are not a blast, at any count: post_outreach_bulk-send sends up to 25 a call, each with its own words. POST /v1/blasts (access: workspace)

post_blasts_id_audience

ChatGPT
Add people to a blast's audience Appends named addresses to a draft blast's audience: recipients as [{address, displayName, sourceRowId, values}], at most 500 a call, where values is that recipient's merge values as {"first_name": "Ana"} — what the blast's {{fields}} are filled from. To aim at everyone on a board who matches a filter, use post_blasts_id_audience_board instead; it resolves the addresses and the values itself. Answers 200 with a result per address — ok and changed for each, and a sentence for any one refused — never a 4xx because a single address was wrong. An address whose shape does not fit the blast's kind (an email in a text blast) is refused there and then instead of failing quietly at send time. Append-only: nobody can be taken out, and nothing can be added once the blast is scheduled or dispatched, so a mis-built list is fixed by cancelling the draft and building another. POST /v1/blasts/:id/audience (access: workspace)

post_blasts_id_audience_board

ChatGPT
Aim a blast at a board Aims a draft blast at the people on a board in ONE call — the way to give a blast its audience. It works on ANY board the caller can open — Leads, Deals, Merchants, Lenders, Agents, Contacts, or one somebody made — as long as it has a column holding phone numbers (text blast) or email addresses (email blast). That need not be a column of the phone or email kind: an imported board often keeps them in a plain text or number column, and addressColumnId may name any text or number column that does. Body: definitionKey (the board, e.g. "leads"), an optional addressColumnId (which phone or email column to send to, when the board has several — absent takes the board's own default), and a filter, at least one of which must pick somebody: the GRID'S filter — q and filters ({column label or id: [values]}, exactly what get_definitions_key_rows and post_definitions_key_rows_act take: a status by its word or tab, any column a computed one included, $ keys; {"filters": {"Documents": ["complete"]}} is everyone not missing docs) — or the older statuses (only on a board that has a status column), tags, search (text found in any cell), rowIds, folded into the same filter. A column the board does not have is refused by name. The service finds the matching rows the caller may see, takes each one's phone (text blast) or email (email blast), leaves out anyone marked do-not-contact, and freezes each recipient's merge values beside the address so the blast's {{fields}} can be filled per person. It answers matched (rows the filter found), skippedNoAddress (of those, the ones with nothing in the address column — counted, never dropped in silence), requested (those with an address for this kind), changed (added by this call), truncated (true past 2,500 a call), noSelection (the request picked nobody — different from matching nobody), only the refused addresses with why, fields — the merge-field names this audience can fill — and addressColumns, the columns it is most likely addressed from, with the default marked. Draft-only and append-only: a mis-aimed list is fixed by cancelling the draft and building another. Then read get_blasts_id_preview. POST /v1/blasts/:id/audience/board (access: workspace)

post_blasts_id_cancel

ChatGPT
Cancel a blast Closes a blast with a reason so it never goes: {"reason": "…"}. Allowed while it is a draft or scheduled only — once dispatch has started, the copies are in the outbox and this answers 409, so withdraw them one at a time with post_comms_messages_id_cancel. There is no route that deletes a blast and no grant for one: that a workspace aimed a message at a list of people outlives the campaign, and cancel is the verb. POST /v1/blasts/:id/cancel (access: workspace)

post_blasts_id_dispatch

ChatGPT
Send a scheduled blast Sends a blast, and is the only verb that does: it expands into one ordinary outbound message per audience address, each in that address's own thread with that recipient's merge fields filled, and the rail draining the outbox carries them. Body: medium (absent takes the blast's kind — "sms" for a text blast, "email" for an email one; a text blast may ask for "imessage" instead, and a crossed pair is refused); scheduleAt (an RFC3339 time, or "" for now — its presence launches a DRAFT in one step, freezing the audience and queueing every copy for that moment); pacing ({spreadMinutes, activeHoursStart, activeHoursEnd, activeHoursTimezone} — spread the copies over that many minutes and only inside those daily "HH:MM" hours in that IANA zone; it also launches a draft, and a pacing that cannot be planned is refused, never ignored); and limit (bounds one run of an already-scheduled blast; addresses past it are ledgered skipped; not allowed with scheduleAt or pacing). A recipient who opted out, or whose copy has a field with nothing to fill it and no fallback, is ledgered skipped with the reason and the rest still go. The copies leave on the blast's sendFrom line, or the workspace's default when it names none; the line is checked again here against what the caller may send from. IDEMPOTENT: a blast already dispatched is answered from its ledger with created: 0, so a retried request cannot message an audience twice. Read get_blasts_id_preview first. A blast is timed by scheduleAt or pacing here, never by a wait in the caller's own loop, and nothing has been sent or scheduled until this or post_blasts_id_schedule has answered. This is the many-people send; post_comms_send is the one-person one. POST /v1/blasts/:id/dispatch (access: workspace)

post_blasts_id_schedule

ChatGPT
Schedule a blast Freezes a draft blast's audience and names the moment it may go: {"at": "<RFC3339 time>"}, absent meaning now. THIS DOES NOT SEND — it closes the audience and writes two columns, and after it neither the words nor the people can change. post_blasts_id_dispatch is what turns a scheduled blast into messages. POST /v1/blasts/:id/schedule (access: workspace)

post_boards

ChatGPT
Make a board Creates a custom board — a table somebody makes for themselves, as opposed to a pack page like leads or deals — in ONE call, choices included. Body: name, an optional purpose, and columns as {label, kind, config}, kind being one of text, number, date, select, checkbox, url, email, phone, person, tags, currency, percent, rating, relation, datetime, long_text, address, file, formula, lookup. A select column takes its choices HERE: config.options as [{label, value?, tone?}] (tone: secondary, brandLight, brand, warning, success, destructive) — never one options call per choice afterwards. A relation (record link) column needs config.definitionKey, the page or board it links to — a page key such as leads, or a board's own board-… key — and is refused without it. Always send recordTabs too: which tabs the board's record pop-up shows besides Overview — send activity, notes, files and tasks on every board, plus calls when the rows carry a phone or email column — fewer only when the person asks for fewer. An unknown tab is a 400 naming it; left out, the app shows its default set. To make a board FROM data instead, reach for post_boards_import and omit its boardId. MANY ITEMS PER ROW (an order's 25 line items, a client's orders, a deal's documents) is NOT a select or a long text: a select holds ONE fixed choice. Make a second board for the items with a relation column pointing back at this one (each item one row, its quantity and status in their own columns), or a child page with subrecords. POST /v1/boards (access: workspace)

post_boards_copy

ChatGPT
Copy a board or a page into a new board Duplicates a custom board or a pack page (leads, deals…) into a NEW board in ONE call — the way to answer "make a copy of X", "a new board with the same columns as X", or "copy X and bring its rows across". Body: from (a board id, or a page key such as leads), name for the new board, an optional purpose (defaults to the source board's), and rows (default true: every row you can read comes across; false copies the columns alone). Columns keep their label, kind, config and choices, in order; a page's status column becomes a select carrying the page's statuses. Rows go through the ordinary row writer, so formula and lookup cells recompute on the copy; file cells do NOT come across (files stay with the source's rows), and a person cell naming someone outside the workspace is left empty — both are listed in skipped with the reason. Answers 201 with the new board, counts {columns, rowsCopied, rowsSkipped} and skipped. A board or page you cannot read is a 404. At most 5000 rows — more is 413 too_many_rows. Never build a copy from post_boards followed by post_boards_import: an import carries only the rows it is sent, and one with none is refused. POST /v1/boards/copy (access: workspace)

post_boards_id_actions

ChatGPT
Add a button to a board's action bar Adds one button to a custom board's action bar: label, icon, kind (add_row, builtin or workflow) and the config that kind reads. Manage on the board. POST /v1/boards/:id/actions (access: board:manage)

post_boards_id_columns

ChatGPT
Add a column to a board Adds one column to a custom board's shape: label, a kind (text, number, currency, date, select, person, file, formula …) and the per-kind config — a select column's choices go in config.options as [{label, value?, tone?}], in this same call; a relation (record link) takes config.definitionKey, which may name any page or any other custom board (its definitionKey, board-…). Rows already on the board keep the cells they had and are simply empty in the new column. Manage on the board; a pack page's column is post_definitions_key_columns. MANY ITEMS PER ROW (an order's 25 line items, a client's orders, a deal's documents) is NOT a select or a long text: a select holds ONE fixed choice. Make a second board for the items with a relation column pointing back at this one (each item one row, its quantity and status in their own columns), or a child page with subrecords. A record's Notes tab draws the column labelled Notes; a board without one gets it here (label Notes, kind long_text). POST /v1/boards/:id/columns (access: board:manage)

post_boards_id_columns_columnId_options

ChatGPT
Add a choice to a board's select column Adds choices to a select column's vocabulary — a whole list in ONE call as options: [{label, value?, tone?}], or one as label, an optional snake_case value (derived from the label when absent) and a tone (secondary, brandLight, brand, warning, success, destructive, or a lowercase hex colour). Send the list, never one call per choice. A board or column being CREATED takes its choices in config.options instead. This is a custom board's vocabulary; a pack page's is its statuses, post_definitions_key_statuses. Manage on the board. POST /v1/boards/:id/columns/:columnId/options (access: board:manage)

post_boards_id_imports

ChatGPT
Record a board import run Record an import somebody ELSE performed row by row, so it lands in the board's history and can be rolled back. POST /v1/boards/import records its own run and is what to reach for to actually import — use this only when the rows have already been written and you are reporting {fileName, total, imported, failed, results}. The row ids in results are stamped as this batch's, which is what makes the undo exact. POST /v1/boards/:id/imports (access: board:write)

post_boards_id_imports_runId_undo

ChatGPT
Undo a board import Roll one import batch back: delete exactly the rows that run CREATED and nothing else — never a row it merely updated, and never a row somebody typed. Answers {run, removed, kept}. Refused with 409 import_running when the run never settled and 409 already_undone when the batch has been rolled back before, because "they were gone before you asked" and "I removed them" are different answers. Get the runId from GET /v1/boards/:id/imports. POST /v1/boards/:id/imports/:runId/undo (access: board:write)

post_boards_id_rows

ChatGPT
Add a row to a board Puts new rows on a custom board: ONE as cells keyed by column name, or SEVERAL in one call as rows: [{cells}, ...] (up to 200, all or nothing, answered as {rows} in the order sent) — seeding sample data or adding a handful of records is one call, never one per row. A cell is keyed by its column's name or its id (trimmed, any case) and a select cell takes its choice's label or value; a label two columns share, a key no column answers to and a word the column has no choice for are each refused naming what there is, never stored as sent. Every value is sent as text; an address cell takes the whole address as one string. For a spreadsheet reach for post_boards_import; for a pack page (leads, deals, contacts) the twin is post_definitions_key_rows. Write access on the board. POST /v1/boards/:id/rows (access: board:write)

post_boards_id_rows_delete

ChatGPT
Delete records from a board Removes the rows named in ids from one custom board, permanently — a board row has no archive to be reactivated out of. Below manage it reaches only the rows the caller can already see. The board itself survives; delete_boards_id is what destroys that. POST /v1/boards/:id/rows/delete (access: board:write)

post_boards_id_rows_rowId_files

ChatGPT
Attach a file to a board record Declares a file against one record on a custom board and answers the row plus a signed upload url; the bytes go straight to storage in a second request and never through this service. WRITE access, not read — pinning a file changes the record. The pack-page twin is post_definitions_key_rows_rowId_files. POST /v1/boards/:id/rows/:rowId/files (access: board:write)

post_boards_id_rows_rowId_transfer

ChatGPT
Duplicate or move a board record safely Copies a custom-board row with a validated column mapping, or moves that same row identity to another board only when no attached data would be detached. A durable idempotency key makes retries return the original result; nonempty unmapped cells and incompatible select values are refused. Many at once: post_boards_id_rows_act POST /v1/boards/:id/rows/:rowId/transfer (access: board:write)

post_boards_id_transfer-preview

ChatGPT
Preview board transfer mappings Compare source and destination custom-board columns before a duplicate or move. It suggests only unique compatible matches and reports unmapped fields and unsupported attachments; no rows are written. Body: targetBoardId (a board id or its board-… key) and an optional mapping {source column id: target column id}. Answers mappings (each with its conversion and confidence), unmapped (each with its reason) and warnings. It reads schemas only: a row's values and linked records are checked when the transfer runs — post_boards_id_rows_rowId_transfer, or the workflow preset record.transfer_board (sourceBoard, sourceRow, targetBoard, mode duplicate or move, mapping) — so a clean preview does not prove a given row will transfer. POST /v1/boards/:id/transfer-preview (access: board:read)

post_boards_id_views

ChatGPT
Save a view on a board Saves one filter, sort and set of visible columns under a name. scope is user (the caller's own, the default) or workspace (everyone's, which takes manage on the board); key is a slug and label is what people see. The pack-page twin is post_definitions_key_views. POST /v1/boards/:id/views (access: board:read)

post_boards_import

ChatGPT
Import rows into a board Put a spreadsheet's worth of rows onto a custom board in ONE call, and the way to make a board FROM data: omit boardId and send name, and the board is created. columns are ENSURED by label (case-insensitive) — a label that already exists is reused, and a kind that disagrees with the existing column is a 400 naming it. Values are coerced per kind server-side: "$1,200.50" into a currency cell, "1/15/2026" into a date, "yes/x/1" into a checkbox, "a, b" into tags, a status label into the option's value. identity names the columns that say which row is which, so a matching row is UPDATED rather than duplicated. Three policies, each defaulting to the first word: unknownChoice add|blank|reject (a status value the column has no option for becomes a new option, an empty cell, or a failed row); duplicates first|last|all (the file names one identity twice); dryRun true|false (plan only, nothing written, answers 200). rows is required — an import with none is a 400; to duplicate a board or page with its rows, use post_boards_copy. At most 5000 rows per call — more is 413 too_many_rows, so page it. Answers the board, the columns and options it created, per-row outcomes (created/updated/skipped/failed with the reason), and a runId the rollback works from. When it MAKES the board, send recordTabs too — which tabs the record pop-up shows besides Overview, activity, notes, files and tasks on every board, plus calls for a phone or email column (fewer only when the person asks); with a boardId it is a 400, since patch_boards_id changes an existing board's tabs. POST /v1/boards/import (access: workspace)

post_boards_import_auto

ChatGPT
Import a spreadsheet onto a board, mapping it for you The custom-board twin of post_definitions_key_import_auto: send a file as headers plus rows and the service works out the mapping itself. In Neo chat, when the person attached a spreadsheet, send attachment with the file's name INSTEAD of headers and rows — the rows are supplied at the door, never retyped. A Google Sheets, OneDrive or CSV link pasted into the chat arrives as an attached file and goes the same way. Pass boardId to import onto a board that exists, or name with no boardId to MAKE a board from the file — which is the call to reach for when somebody wants a spreadsheet turned into a board, because the columns and their kinds are read off the file's own values. The answer carries the plan it used — what each header became, why, and how sure it was — beside the ordinary import result. dryRun: true is for a caller who wants to check a mapping before anything is written — the plan still comes back; a file attached in Neo chat is imported, never dry-run, because the file in the chat is the instruction. A new dropdown arrives with every choice the file uses, ordered, coloured and grouped, and the answer's columnsCreated and optionsAdded say what was made. At most 5000 rows, refused whole with too_many_rows, and post_boards_id_imports_runId_undo takes the run back. When it MAKES the board, send recordTabs too — which tabs the record pop-up shows besides Overview, activity, notes, files and tasks on every board, plus calls for a phone or email column (fewer only when the person asks); with a boardId it is a 400, since patch_boards_id changes an existing board's tabs. POST /v1/boards/import/auto (access: workspace)

post_boards_import_cells

ChatGPT
Read the cells an import column could not, with the model The importer's last resort for ONE column: send header, the kind of the column the cells are going into (number, currency, percent, date, checkbox, select, text…), its choices when it is a select, and cells — each {row, value, beside} where beside is the row's other cells by header for context. Answers a value per row in the column's own form (plain digits, an ISO date, true/false, one of the choices spelled as given) or an empty string to leave it blank, with a confidence and a reason; every value is checked against the kind and anything the column could not hold comes back blank. Nothing is written — put the readings into the rows you then send to post_boards_import. At most 200 cells per call. With no model connected the answer is 200 with no cells and modelProblem saying why. POST /v1/boards/import/cells (access: workspace)

post_calendar_events

ChatGPT
Book a meeting Puts a meeting on the caller's own calendar (Google or Microsoft, whichever they connected) and links it to a record. Takes title, startsAt (required), an optional endsAt (the member's default duration when absent) — both RFC3339 instants WITH an offset, such as 2026-10-06T15:00:00-07:00, written in the person's own time zone (the clock line names it); a bare date or a time with no offset is refused — attendees, meetLink for a video link, and definitionKey + rowId to pin it to a record. attendeesFromRecord adds the record's own addresses and the contacts filed against it; attendeesFromAssignees adds the people its person columns name (opener, closer, owner) — so nobody retypes what the record already knows. To tie a task: taskId links an existing task of this record (refused if it already has a live meeting), or createTask {title, description, dueAt, assigneeAccountId, entityKind} raises one beside the meeting — blanks default to the meeting's title, its start as the due time, the caller as the assignee, and the record's own entity word; the answer carries taskId. Never send both. This is a write: it pauses for the person's approval, which is where they accept or decline the invite before it goes out. When the meeting ALREADY EXISTS on their calendar, do not book another: post_calendar_events_link puts that one on the record. POST /v1/calendar/events (access: workspace)

post_calendar_events_link

ChatGPT
Put an existing meeting on a record Files a meeting the person ALREADY HAS on their own calendar on a record, instead of booking a second one. Body: providerEventId (the entry's providerEventId from get_calendar_events with from/to), definitionKey + rowId (the record), and account to read one connected calendar only (left out, each connected calendar is tried in turn, past one that needs reconnecting). The calendar is only read: the meeting stays the person's, so moving or renaming it is done in Google or Outlook (patch_calendar_events_id refuses a linked meeting) and delete_calendar_events_id takes it OFF THE RECORD without touching their calendar. Linking the same meeting to the same record again answers the existing link; one already on another record, or on a colleague's, is a 409. POST /v1/calendar/events/link (access: workspace)

post_calendar_meet-sync

ChatGPT
Enable Google Meet transcript tasks Enrolls the integration administrator's own Google connection for event-driven Drive notifications and a one-time 30-day notes/transcript backfill. Requires connectionId, definitionKeys ordered business records before contacts, and an IANA timezone. Uses existing Drive read-only and Calendar grants and workspace Neo configuration. Creates evidence-backed tasks only for explicit internal commitments; ambiguous or older meetings remain in review. Re-saving a source retries its review documents after correcting record links. Does not send mail or invitations. POST /v1/calendar/meet-sync (access: workspace)

post_cash-flows

ChatGPT
Create a cash flow Records a movement of money expected on a deal — an instalment coming in, a disbursement going out, a fee or a refund. Body: dueDate (required; a cash flow needs its due date) and expectedAmountMinor (more than zero), plus flowType (payment, disbursement, commission, fee, refund, adjustment), direction (inbound or outbound), method (unspecified, ach, wire, check, card, revenue_split), currency, scheduleId, facilityId, dealRowId (the deal record on our own board), counterpartyId, sequenceNumber, totalPayments and notes. It lands at expected: money that then arrives is post_cash-flows_id_record-payment. POST /v1/cash-flows (access: workspace)

post_cash-flows_id_mark-failed

ChatGPT
Mark a cash flow failed Records that the attempt to move the money failed. Body: reason, required — the ledger refuses a failed flow that does not say why — and code, the processor's own code, optional. Legal only from the unsettled states, and a failed flow can still take a receipt later if the money turns up (post_cash-flows_id_record-payment). Money that simply never arrived, with no failed attempt behind it, is post_cash-flows_id_mark-missed. POST /v1/cash-flows/:id/mark-failed (access: workspace)

post_cash-flows_id_mark-missed

ChatGPT
Mark a cash flow missed Records that the money never arrived: the flow's status becomes missed. Body: reason, optional. Legal only from the unsettled states. A collection that was ATTEMPTED and failed is post_cash-flows_id_mark-failed instead, which demands its reason. POST /v1/cash-flows/:id/mark-missed (access: workspace)

post_cash-flows_id_record-payment

ChatGPT
Record a payment against a cash flow Records that the money on a cash flow actually arrived: the status becomes completed and the receipt is stored against it. Body: actualAmountMinor (required and more than zero — a receipt needs the amount that actually landed), processedDate (today if absent), externalTransactionId, batchId, notes. Legal only from expected, pending or failed; an already-settled flow reads as a not-found. A payment schedule's instalment is a different record — that receipt is post_payments_id_record. POST /v1/cash-flows/:id/record-payment (access: workspace)

post_comms_calls_dial

ChatGPT
Place a call from a record Dial one of a record's phone numbers through the workspace's connected dialer. The call lands in the call log immediately as an outbound entry, and the dialer's own status callback moves it from ringing to connected to ended with a duration. A number that asked us to stop is refused with 409 and the same sentence a text gets. Body: definitionId and rowId (the record — both required, the call is filed under it) and phone (required — the number to dial, one of that record's phone cells; read the record first). It needs the workspace's Orum dialer connected (Settings › Integrations): without it every dial answers 422 "its Orum integration is not connected" — say so and stop, a different record will not change it. POST /v1/comms/calls/dial (access: workspace)

post_comms_check

ChatGPT
Check addresses before reaching out Checks up to 100 email addresses or phone numbers before anything is sent, in the order given, and sends, spends and changes nothing. Body {addresses: [..], medium: email|sms|imessage|call (optional)}. Each result has the address in its one spelling (a mailbox lower-cased, a number as E.164), its kind, a verdict (pass, warn, fail or unknown — the worst of its checks) and the checks, each with an outcome and a reason in words: email_format, email_typo, email_role (a shared mailbox like info@), email_disposable, email_domain (whether the domain takes mail), phone_format, phone_in_service, phone_mobile (from mobile lookups already made: unknown until a text is sent), do_not_contact, and medium (whether it fits the medium asked about). Use it before a send, a blast or a list of outreach to say who can be reached and why not; a fail on do_not_contact means every send to it is refused. POST /v1/comms/check (access: workspace)

post_comms_connections_id_aliases_import

ChatGPT
Import a Gmail account's aliases and signatures Reads the send-as list Gmail holds for one of the caller's OWN connected Google accounts and makes the workspace agree with it: every alias Gmail has accepted becomes an inbox the caller owns, an alias Gmail no longer lists is removed, and the signature Gmail keeps for the account or an alias is copied to that address — never over a signature somebody typed here. Answers the aliases as they now stand, how many were added and removed, how many signatures were copied, and each address it left out with the reason (not verified in Gmail, already connected as its own account, a typed signature kept). Safe to run again. Put people on an alias afterwards with put_comms_inboxes_medium_identifier_members_accountId. An Outlook account is refused, and Gmail's own refusal comes back as a 502 in Gmail's words. POST /v1/comms/connections/:id/aliases/import (access: workspace)

post_comms_conversations

ChatGPT
Open a conversation with an address Opens the thread with one address — a medium ("sms", "imessage", "email") and the contact_identifier — or answers the one that already exists. Idempotent on the address, so a caller that cannot remember whether it has written before can simply ask again. Nothing is sent here and nothing names a record. POST /v1/comms/conversations (access: workspace)

post_comms_conversations_id_calls

ChatGPT
Log a call that already happened Records a call on a thread: direction, who, when it started, how long it ran, the result, and a recording URL, transcript or summary if there is one. Nothing is dialled — a call is RECORDED here. To place one, use POST /v1/comms/calls/dial. POST /v1/comms/conversations/:id/calls (access: workspace)

post_comms_conversations_id_messages

ChatGPT
Send a message on a conversation Queues a message on a conversation that already exists — the reply. The medium comes from the thread, so the one route sends a text or an email; subject and content_html belong to an email thread and are refused on a text. Reach for this when you have the conversation id, and for POST /v1/comms/send when you have only an address: answering a thread a read handed you is a reply here, never a second thread opened with the send. template_id (with definition_key + row_id for the record that fills it) sends a workspace template the same way POST /v1/comms/send does. POST /v1/comms/conversations/:id/messages (access: workspace)

post_comms_conversations_id_messages_preview

ChatGPT
Preview a message on a conversation What post_comms_conversations_id_messages would do with the SAME body, and nothing is sent or saved: medium, from (our line it leaves on, as the thread will carry it; "" is no particular line), the other side (contact_identifier, contact_name), conversation_id, parent_message_id (the earlier message an email is threaded under, null for a new one), the words as they will go (subject, body, body_html, with any template_id filled) and signature (the HTML the send appends, "" for none — a text is never signed). Decided by the send's own rule, so the send then does exactly this. Whatever that send would refuse is refused here in the same words. To send it, call post_comms_conversations_id_messages itself. POST /v1/comms/conversations/:id/messages/preview (access: workspace)

post_comms_conversations_id_mute

ChatGPT
Mute a conversation Silences one thread for the CALLER and nobody else. {"is_muted": false} unmutes it. POST /v1/comms/conversations/:id/mute (access: workspace)

post_comms_conversations_id_participants

ChatGPT
Add somebody to a conversation Puts a colleague (user_id) or an outside contact (external_contact_name with an email or a phone) on a thread's roster, with an optional role. POST /v1/comms/conversations/:id/participants (access: workspace)

post_comms_conversations_id_read

ChatGPT
Mark a conversation read Clears the caller's unread badge on one thread. Unread is per person, so this never clears a colleague's. POST /v1/comms/conversations/:id/read (access: workspace)

post_comms_conversations_id_records

ChatGPT
File a conversation under a record Links one thread to a record so it shows on that record's Communications tab: row_id plus the page as definition_key (its key, like "leads") or definition_id. Pure metadata: one row, and not a single message is touched. Not needed after post_comms_send that named the record — that send files its own thread. POST /v1/comms/conversations/:id/records (access: workspace)

post_comms_drafts_ai

ChatGPT
Draft a message with Neo Writes a first pass at an email or text about one record and files it in the approval queue, unsent. Takes the record (definitionKey, rowId), a medium (email or sms), an intent saying what the message is for, the recipient, an optional tone (shorter, warmer or firmer), an optional note in the person's own words, and an optional templateId to rewrite from. The answer STREAMS as server-sent events and ends with the draft row's id, its subject and body, and any figure in the words the record's own facts do not support. It never sends: approving the draft and sending it are separate acts. The draft names no mailbox and no From can be set on it: to send from a line the person named, send the words with post_comms_send and account_identifier set to it from get_comms_senders. THE MODEL WRITES THE WORDS: note and intent steer it, and it may reword, shorten or add to them — so it is the wrong tool when the person gave the exact wording ("use exactly this", a body pasted in quotes). For that, send the words as given (post_comms_send, or post_comms_conversations_id_messages to answer a thread), shown first in the plan when the person wants to review them; post_message-drafts with definitionKey + rowId saves the words as given only when they ask for a draft to approve later (2026-10-08: dictated messages came back with figures and a misspelled name before being patched). On a record, the draft also reads the person's standing rules and that record's newest thread of the medium, our messages included; conversationId beside the record names which of its threads it answers. An email draft carries no sign-off, because its sending address signs it; a text draft ends as the person's rule for texts says, or without one. POST /v1/comms/drafts/ai (access: workspace)

post_comms_labels

ChatGPT
Create an inbox label Makes a workspace inbox label. Body: name (1 to 60 characters, unique in the workspace whatever its case) and optional tone (brand, brandLight, warning, success, secondary, destructive). Any seat may make one. POST /v1/comms/labels (access: workspace)

post_comms_messages_id_cancel

ChatGPT
Cancel a queued message Withdraws a message that has not gone out yet. Only while it is still queued — once a sender holds it we cannot know whether it has already left, and that comes back as a not-found rather than a cheerful yes. POST /v1/comms/messages/:id/cancel (access: workspace)

post_comms_numbers

ChatGPT
Buy a phone number Provision one number the search offered onto the workspace's carrier account and record it as a line the workspace holds. It then becomes selectable as the sending number. This costs the workspace money every month. POST /v1/comms/numbers (access: workspace:admin)

post_comms_opt-outs

ChatGPT
Add an address to the do-not-contact list Lists one phone number or email address as do not contact, as the caller, with an optional reason; every later text, email, call and blast copy to it is refused. Body {address, reason}. The answer is the line in its one spelling (a number as E.164, a mailbox lower-cased); an address already listed answers the line that stands. Anything that is not a phone number or an email address is a 400. Use it when a person asks, in any words or on a call, not to be contacted. POST /v1/comms/opt-outs (access: workspace)

post_comms_recordings_id_summarise

ChatGPT
Summarise a call recording Write the recordings summary from its transcript, replacing any summary already stored and keeping the transcript. Refuses with 409 when there is no transcript yet. POST /v1/comms/recordings/:id/summarise (access: workspace)

post_comms_records_definitionId_rowId_read

ChatGPT
Mark a record's conversations read Clears the caller's unread badges on every thread filed under one record, in one sweep, and answers how many moved. Zero is a normal answer. POST /v1/comms/records/:definitionId/:rowId/read (access: workspace)

post_comms_send

ChatGPT
Send a text or an email to an address THE SEND: a medium ("sms", "imessage" or "email"), the address to write to (contact_identifier) and the words (content). Opens the thread with that address if this workspace has never written to it, then queues the message. No record is required and no conversation id — reach for this to compose to somebody: "send an email to …", "email this person", "send them a text", "text this number". When the send is ABOUT a record (a lead, a deal), name it with definition_key + row_id: the thread is filed on that record, so it shows in the record's communications and the answer's conversation.records lists it — no separate link call. With no record named, a record whose phone or email cell holds the address files it; conversation.records says where it landed, and empty means nowhere. An email also takes subject and an optional content_html. NOTHING refuses an email with no subject and nothing supplies one, so it arrives blank in somebody's inbox — on 2026-09-20 a model asked to "send an email that says hi" sent exactly that. Write a subject for an email unless the person gave you words that are plainly the subject. An address that asked us to stop is refused. Name template_id to send a workspace template instead of writing the words: definition_key + row_id name the record whose values fill its {{placeholders}} (read under your own row wall and column mask). A placeholder the record has no value for refuses the send (400, naming it); an email template on a text, or the reverse, is refused; a subject or content you also send wins over the template's. account_identifier is the line it leaves on, spelled as get_comms_senders lists it — name it whenever the person named one; left out, a text leaves on the caller's OWN line when they have one (own: true in get_comms_senders), otherwise the workspace's default line. A line the caller may not send from is refused (line_refused). Many at once: post_outreach_bulk-send POST /v1/comms/send (access: workspace)

post_comms_send_preview

ChatGPT
Preview a send to an address What post_comms_send would do with the SAME body, and nothing is sent or saved: medium, from (our line it leaves on — the one named, else the one the send would choose; "" is no particular line), who it reaches (contact_identifier, contact_name), conversation_id (the thread it lands on, null when the send would open one), parent_message_id (the message an email goes as a reply to — a mail to an address the workspace already writes to replies on that thread), the words as they will go (subject, body, body_html, with any template_id filled) and signature (the HTML the send appends, "" for none — a text is never signed). Decided by the send's own rule, so the send then does exactly this. Whatever the send would refuse — a line the caller may not send from, a template field with no value, an email with no words — is refused here in the same words. POST /v1/comms/send/preview (access: workspace)

post_credit-memos

ChatGPT
Create a credit memo Opens a credit memo as a draft — the underwriting write-up on a deal. Body: title (required; the ledger refuses a blank one), memoType (initial, renewal, modification, exception), facilityId, riskProfileId, previousMemoId (the memo this one supersedes), the nine narrative panes as top-level fields (executiveSummary, businessOverview, financialAnalysis, industryAnalysis, riskAssessment, collateralAnalysis, ownershipAnalysis, strengthsWeaknesses, recommendation), currency, requestedAmountMinor, recommendedAmountMinor, recommendedRate, recommendedTermMonths, recommendedProduct, recommendedDecision, recommendedConditions, attachmentIds, expiresAt, aiGenerated and generationModel. It lands as a draft; sending it up for a decision is post_credit-memos_id_submit. POST /v1/credit-memos (access: workspace)

post_definitions_key_actions

ChatGPT
Add a row action to a page Adds a button to a record page's action bar. kind is add_row (the default), builtin or workflow, label is up to 60 characters, and icon and config carry what the kind needs — a workflow action's config names the workflow it runs. The new button goes last on the bar. Workspace admin. POST /v1/definitions/:key/actions (access: def:admin)

post_definitions_key_columns

ChatGPT
Add a column to a page Adds a column to a record page. kind is one of text, number, date, select, checkbox, url, email, phone, person, tags, currency, percent, rating, relation, datetime, long_text, address, file, formula, lookup, and config is validated against that kind — a select's options, a relation's target (config.definitionKey: any page OR any custom board, whose key is the board's definitionKey, board-…), a formula or lookup's references, which are checked against columns that actually exist. source defaults to own and otherwise must read crm:<object>.<field> or app:<object>.<field>; priority defaults to 50 (0-1000); minWidth is 40-800, or 0 (the default) to let the app size the column by its kind and header. The new column goes last in the page's order. Workspace admin. A record's Notes tab draws the column labelled Notes; a page without one gets it here (label Notes, kind long_text). POST /v1/definitions/:key/columns (access: def:admin)

post_definitions_key_columns_columnId_options

ChatGPT
Add choices to a page's select column Adds choices to a select column's vocabulary on a record page — a whole list in ONE call as options: [{label, value?, tone?}], or one as label, an optional snake_case value (derived from the label when absent) and a tone (secondary, brandLight, brand, warning, success, destructive, or a lowercase hex colour). Send the list, never one call per choice. The board door's twin; workspace admin. POST /v1/definitions/:key/columns/:columnId/options (access: def:admin)

post_definitions_key_entity-groups

ChatGPT
Name a relationship Makes an EMPTY entity group on this page — a named relationship with a kind (affiliated, parent_subsidiary, common_ownership, franchise, successor). Members are added afterwards. Use it when two or more records on one page turn out to be the same commercial party. POST /v1/definitions/:key/entity-groups (access: def:read)

post_definitions_key_entity-groups_groupId_members

ChatGPT
Put records into a relationship Adds up to 100 rows to one entity group. Answers 200 with a line per row: membership is EXCLUSIVE, so a record already in another relationship is refused individually with the sentence saying so while the rest of the batch lands. POST /v1/definitions/:key/entity-groups/:groupId/members (access: def:read)

post_definitions_key_groups

ChatGPT
Create a record group on a page Makes an EMPTY record group on a page, from name (required), description and colour — that it can be empty is the whole difference from a status. It lands last in the board's order. A name already used on this page is a 409, and the check is case-blind, so 'Q4 Push' and 'q4 push' collide. Put records in with post_definitions_key_groups_groupId_rows. POST /v1/definitions/:key/groups (access: def:read)

post_definitions_key_groups_groupId_rows

ChatGPT
Add records to a group Puts the records named in ids into one group, in a single transaction, with a line per record saying what happened. A record already in the group is a no-op rather than an error, and a record that is not on this page is refused on its own line while the batch carries on. At most 100 ids. POST /v1/definitions/:key/groups/:groupId/rows (access: def:read)

post_definitions_key_groups_groupId_rows_remove

ChatGPT
Remove records from a group Takes the records named in ids out of ONE group, with a line per record. The records are NOT deleted — that is the point of a membership — and a record that was not in the group is a no-op. A POST rather than a DELETE because a hundred ids do not belong in a URL. At most 100 ids. POST /v1/definitions/:key/groups/:groupId/rows/remove (access: def:read)

post_definitions_key_groups_move

ChatGPT
Move records between groups Moves the records named in ids out of the from group and into the to group in ONE transaction, both groups on this page. A record that is not in the source group is REFUSED on its own line rather than quietly added to the target — 'move' is a claim about where it was. The same group for both is refused outright. At most 100 ids. POST /v1/definitions/:key/groups/move (access: def:read)

post_definitions_key_groups_rows_remove

ChatGPT
Take records out of their groups Takes the records named in ids out of EVERY group on this page in one transaction — 'in no group' said once, instead of reading the group list and calling remove per group. A record in three groups leaves all three and files one timeline line per membership actually ended; a record in no group at all is a no-op. At most 100 ids. It names no group, which is why this path has no groupId. POST /v1/definitions/:key/groups/rows/remove (access: def:read)

post_definitions_key_import

ChatGPT
Import rows into a page Imports up to 5000 rows into a pack page in one call, coercing each cell to its column's kind — amounts, dates, checkboxes, tags — and ensuring by label any column the file needs. Above 5000 rows the call is refused whole, so send it in pages. identity names the columns that say which record is which: a row that matches one already on the page is merged, not appended. unknownChoice decides a select value the column does not carry (add | blank | reject), duplicates decides a record the file names twice (first | last | all), and dryRun plans the whole thing without writing. A row that cannot be written fails on its own line and the run carries on; the run is recorded, so the whole batch can be undone. One at a time: post_definitions_key_rows POST /v1/definitions/:key/import (access: def:read)

post_definitions_key_import_auto

ChatGPT
Import a spreadsheet, mapping it for you Imports a file onto a pack page WITHOUT being told which column is which — send the sheet as it came: headers (its first row) and rows (every other row, as arrays of text in header order). Reach for this whenever somebody hands you a spreadsheet, a CSV or a pasted table for an existing page; reach for post_definitions_key_import instead only when you already know the mapping and have rows keyed by column. The service folds each header onto the page's own columns, reads what every column actually holds, works out which headers identify a record so a re-import updates instead of duplicating, and creates the columns the page is missing. The answer carries the plan it used — one entry per header saying what it decided, why, and how sure it was — beside the ordinary import result, so you can tell the person what went where. dryRun: true is for a caller who wants to check a mapping before anything is written: the same plan and per-row outcomes come back. A file attached in Neo chat is imported, never dry-run — the file in the chat is the instruction. A new dropdown arrives with every choice the file uses, ordered, coloured and grouped, and the answer's columnsCreated and optionsAdded say what was made. Same ceiling as every import — 5000 rows, refused whole with too_many_rows — and the same undo: post_definitions_key_imports_runId_undo rolls the whole run back using the runId in the result. In Neo chat, a spreadsheet attached to the message — a pasted Google Sheets, OneDrive or CSV link arrives as one — goes as attachment with the file's name INSTEAD of headers and rows; the rows are supplied at the door, never retyped. POST /v1/definitions/:key/import/auto (access: def:read)

post_definitions_key_import_jobs

ChatGPT
Import rows as a job Starts the same import as post_definitions_key_import but as a JOB: it writes the run, lands its first chunk of 200 rows inside this call, and answers 202 — completed when the file fitted in that one chunk, running otherwise. Reach for it over the plain import when the file is large enough that somebody is watching, or when a retry has to be safe — send the same idempotencyKey and a second call answers the SAME job instead of landing the file twice. Same body: rows already keyed by column, columns to ensure by label, identity to merge onto records that already exist, unknownChoice (add | blank | reject), duplicates (first | last | all). Still at most 5000 rows, refused whole with too_many_rows above it. If the answer says running, poll get_definitions_key_import_jobs_jobId about once a second — each read lands the next chunk, so 5000 rows takes about 25 reads — until state is completed or failed; the whole job is one batch, so post_definitions_key_imports_runId_undo rolls all of it back. POST /v1/definitions/:key/import/jobs (access: def:read)

post_definitions_key_imports

ChatGPT
Record the results of a completed import Records one FINISHED import in a page's history — fileName, total, imported, failed, an error if it went wrong, and the per-row results. It imports NOTHING: this is the ledger entry for an import the caller already ran itself. To actually import rows use post_definitions_key_import, post_definitions_key_import_auto (a spreadsheet mapped for you) or post_definitions_key_import_jobs. Counts that do not add up — imported plus failed exceeding total — are refused. POST /v1/definitions/:key/imports (access: def:read)

post_definitions_key_imports_runId_undo

ChatGPT
Undo an import batch Removes exactly the rows one import run created and nothing else, keeping any that have since been converted into deals and naming each one it kept. Refuses a run that never finished, and refuses a second undo of the same batch. Reach for it after a mis-import instead of deleting rows by hand. POST /v1/definitions/:key/imports/:runId/undo (access: def:read)

post_definitions_key_queues_queue_pull

ChatGPT
Pull the next record from a work queue Claims the oldest unclaimed deal in one queue FOR THE CALLER and answers it, so two people working the same queue never take the same deal. The queue segment names a stage band of the deals board's own status vocabulary, and an unknown one is refused with the list of bands that exist. role is optional and defaults to opener; unclaimed means that role's person cell is empty. An empty queue is a 404 — 'nothing waiting' is the answer, not a failure. Deals page only. POST /v1/definitions/:key/queues/:queue/pull (access: def:read)

post_definitions_key_rows

ChatGPT
Create a record on a page (lead, deal, merchant, lender, contact, task) Creates one record on a page — a lead, a deal, a merchant, a lender, a contact, a task, an offer, a submission, or a row on any pack page by its key — from cells, each keyed by its column's name or its id (trimmed, any case); a select or status cell takes its choice's label or value; a label two columns share, a key no column answers to and a word the column has no choice for are each refused naming what there is, never stored as sent. The row lands last in the page's order and comes back whole. On the LEADS page this same call is the whole lead intake in one transaction: the duplicate check, the page's default status and the owner. The four money boards (payments, payment-schedules, cash-flows, obligations) refuse — their rows are the ledger's and are written by a lifecycle action, never typed. Many at once: post_definitions_key_import POST /v1/definitions/:key/rows (access: def:read)

post_definitions_key_rows_adopt

ChatGPT
Adopt a shared record onto this page Adopt lenders from the platform catalogue onto the workspace's Lenders board: each catalogue id becomes a directory row carrying the name, address, programs, states and amount range, linked so a retirement shows on it. Idempotent per entry. Admin only. POST /v1/definitions/:key/rows/adopt (access: def:read)

post_definitions_key_rows_assignees

ChatGPT
Assign one person to many records Puts ONE person in ONE role across many records in a single transaction — the bulk form of put_definitions_key_rows_rowId_assignee_role. It works on any page that has assignable people (deals, leads, submissions, merchants) and the page comes from the route, so the body carries no entity type. At most 100 ids; a bad role or a person holding no seat refuses the whole request, while a row id that names nothing is that row's own outcome and the batch carries on. One at a time: put_definitions_key_rows_rowId_assignee_role POST /v1/definitions/:key/rows/assignees (access: def:read)

post_definitions_key_rows_assignments

ChatGPT
Add or remove people on many records Adds the people in add and takes off the people in remove, in ONE role, across up to 100 records named by ids, in one transaction — everyone else on each record stays. Use it to put a second person on a batch of leads or deals without replacing who is already there (post_definitions_key_rows_assignees replaces the role with one person). Each record's person cell keeps its primary unless that person is removed, in which case it moves to whoever remains. Always 200 for a well-formed batch; a record the store refused is reported beside its id. The same account in add and remove is refused. One at a time: post_definitions_key_rows_rowId_assignments_role_accountId POST /v1/definitions/:key/rows/assignments (access: def:read)

post_definitions_key_rows_bulk

ChatGPT
Update many records at once One operation over many deals in ONE transaction: op is status (with status, stage, force, reason), assign (with role and assignee, an empty assignee clearing the role) or delete. At most 100 ids, refused whole above that. The answer is always 200 when the batch itself was well formed, with a line per row saying what landed and why anything was refused — a row the rules turned down is an outcome, not a failed request. Deals page only. It runs through post_definitions_key_rows_act (the same writers; there ids takes thousands and any page) — prefer that. POST /v1/definitions/:key/rows/bulk (access: def:read)

post_definitions_key_rows_delete

ChatGPT
Delete records from a page Destroys the records named in ids and everything that hung off them — timeline, files, documents — and answers how many actually went. Reach for post_definitions_key_rows_rowId_archive instead whenever a deal or lead is simply FINISHED with: archiving keeps the whole trail and can be reversed, and this cannot. Ids from another page, another workspace, or beyond this seat's reach match nothing and are not counted. POST /v1/definitions/:key/rows/delete (access: def:read)

post_definitions_key_rows_intake

ChatGPT
Create a merchant and deal from intake details The composite deal intake: finds or creates the merchant, creates the deal on it, and optionally links the lead it came from — all in ONE transaction, so a failed deal never leaves an orphan merchant behind. merchant.legalName is required; deal.productType, deal.requestedAmount and deal.name fill the new record, and leadRowId writes the relation back to the lead. Answers 201. Deals page only. POST /v1/definitions/:key/rows/intake (access: def:read)

post_definitions_key_rows_onboard

ChatGPT
Onboard a lead Creates a lead, puts its first task on somebody and sends an intro text or email, in one call. The lead and the task are atomic; the greeting is attempted afterwards and its outcome is reported in intro rather than failing the call. Use it instead of three separate calls when taking on a new lead end to end. NEVER for a person already on a page — that makes a second lead (capability sweep, 2026-10-20: "send the application to Dmitri Salas" onboarded him again); a record that exists is sent its form with the record's form-link. POST /v1/definitions/:key/rows/onboard (access: def:read)

post_definitions_key_rows_rowId_account-number_reveal

ChatGPT
Reveal an authorized account number Explicitly reveal the stored account number for this authorized record. Returns available false with a reason when the original number was not retained; never reconstructs a number from its mask. Payout methods are restricted to their owner. POST /v1/definitions/:key/rows/:rowId/account-number/reveal (access: def:read)

post_definitions_key_rows_rowId_affiliate-invitations

ChatGPT
Invite an outside affiliate onto a record Mints a link for somebody outside the firm onto ONE record, under an affiliate preset (rolePresetId, e.g. Medical provider on a Treatment row), to email. They read only the fields the preset lists, the case's identity, and the files it allows; they upload the document types it names, which land on this record. Answers {invitation, token, path, inviteUrl} — the token exists only here. The caller must see the record and be able to edit its board. 400 when the preset is a seat's role or does not cover this board. POST /v1/definitions/:key/rows/:rowId/affiliate-invitations (access: def:read)

post_definitions_key_rows_rowId_ai_insights

ChatGPT
Regenerate a deal's AI insights Reads the deal again — its cells plus whatever the document analysis found — and REPLACES the stored summary and bullets. Spends the workspace's OpenAI budget, so it is an act rather than something a page load does: reach for it when the deal has changed materially, not to refresh a view. Refuses with a sentence when the workspace has connected no OpenAI key, and answers 409 when the deal carries too little to say anything about. POST /v1/definitions/:key/rows/:rowId/ai/insights (access: def:read)

post_definitions_key_rows_rowId_ai_research

ChatGPT
Run business research on a deal Opens a business-research file on the company behind the deal and files it as an ordinary business_research credit record, so GET .../credit-records/business_research lists it immediately. The model has no ability to look anything up, so what it writes is a FIRST PASS over the deal's own material with every unsupported field left empty — the row is filed as working, never done, and the notes say what still has to be checked. Spends the workspace's OpenAI budget; refuses with a sentence when there is no key, and answers 409 when the deal names no business. POST /v1/definitions/:key/rows/:rowId/ai/research (access: def:read)

post_definitions_key_rows_rowId_archive

ChatGPT
Archive a record Takes one record off the live board and keeps it: its cells, its timeline, its files and its documents all stay, and POST .../reactivate puts it back in the position it left from. Reach for this, not rows/delete, whenever a deal or lead is finished with rather than a mistake — the delete destroys the record's whole trail. Already-archived is not found. Many at once: post_definitions_key_rows_act POST /v1/definitions/:key/rows/:rowId/archive (access: def:read)

post_definitions_key_rows_rowId_assignee_role_email

ChatGPT
Email whoever holds a role on a deal Queues an email to the member in one role on one DEAL — opener (the agent) or closer — read off the deal when the call is made, so a workflow written before a reassignment reaches the person who holds the deal now. content is required, subject and content_html optional, account_identifier names the sending inbox (the caller's default when empty). Goes through the same outbox and sending-inbox check as POST /v1/comms/send and answers the same {conversation, message}. A role nobody holds is a 400 naming the role; the caller must be able to see the row and the role's column. Deals only. template_id sends a workspace email template filled from this deal instead of content. POST /v1/definitions/:key/rows/:rowId/assignee/:role/email (access: def:read)

post_definitions_key_rows_rowId_assignee_role_email_preview

ChatGPT
Preview the email to a deal's role holder What post_definitions_key_rows_rowId_assignee_role_email would do with the SAME body, and nothing is sent or saved: who holds the role now (contact_identifier), the mailbox it leaves from (from, "" when none can be chosen), conversation_id, parent_message_id (the message it goes as a reply to, null for a new mail), the words with any template_id filled from this deal (subject, body, body_html) and signature (the HTML appended, "" for none). Decided by the send's own rule; whatever the send would refuse is refused here in the same words. The holder is read when the email is actually sent, so a reassignment in between changes who it reaches. POST /v1/definitions/:key/rows/:rowId/assignee/:role/email/preview (access: def:read)

post_definitions_key_rows_rowId_assignments_role_accountId

ChatGPT
Put one person on a role on a record Adds ONE account to ONE role on a record, keeping everyone the role already holds, and answers the diff with the set as it now stands. Reach for it to put a colleague beside whoever is already on a deal, lead or made-board record — put_definitions_key_rows_rowId_assignments instead replaces the whole set with the list it is given. The record's cell keeps the primary it names; on a role nobody held, the account added becomes the primary. Naming someone already on is a no-op. Roles are agent, closer, originating_agent and processing_agent on deals, owner on leads, submissions, merchants and made boards. Many at once: post_definitions_key_rows_assignments POST /v1/definitions/:key/rows/:rowId/assignments/:role/:accountId (access: def:read)

post_definitions_key_rows_rowId_books_accounts

ChatGPT
Add an account to a client's chart Adds one account: code (unique in this chart; 409 if taken), name, and kind (asset, liability, equity, income, cogs, expense). Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. POST /v1/definitions/:key/rows/:rowId/books/accounts (access: def:read)

post_definitions_key_rows_rowId_books_entries

ChatGPT
Post a journal entry to a client's books Posts one entry: on (YYYY-MM-DD), kind (opening or manual), memo, and 2 to 50 lines of accountId and amountMinor (cents, debits positive, credits negative) that sum to zero - an unbalanced entry is refused. Use it for opening balances and anything the bank feed never sees. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. POST /v1/definitions/:key/rows/:rowId/books/entries (access: def:read)

post_definitions_key_rows_rowId_books_fill-close

ChatGPT
Fill a close from its business's books On a Bookkeeping close: reads its Business and period start and end, builds that business's report for the period, and writes the close's revenue, cost of goods sold, gross profit, operating expenses, net income, cash, total assets, total liabilities, equity and uncategorized count through the row's own cell door (column locks, record wall and undo apply). Refused with what is missing when the close names no business or period, or the business has no books. Answers the report, the cells written and the row. POST /v1/definitions/:key/rows/:rowId/books/fill-close (access: def:read)

post_definitions_key_rows_rowId_books_report_pdf

ChatGPT
File a client's financial statements as a PDF Draws the record's P&L for from..to and balance sheet at to as one PDF and files it on the record as financials; answers the file, which is downloaded through the file's own routes. Never sent to the processor. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. POST /v1/definitions/:key/rows/:rowId/books/report/pdf (access: def:read)

post_definitions_key_rows_rowId_books_setup

ChatGPT
Set up a client's books Seeds the record's chart of accounts and category rules from the small-business chart (a Businesses record) or the household chart (a Clients record), and charts every linked bank account. Running it again adds only what is missing and changes nothing a person changed. Answers the books. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. POST /v1/definitions/:key/rows/:rowId/books/setup (access: def:read)

post_definitions_key_rows_rowId_close

ChatGPT
Close a deal End a deal in one act: move it to a terminal status, record whether it was won, lost or declined with the reason and a closing note, and create the follow-up task — all in one transaction, so a refused half leaves the deal where it was. A decline must say why, and a decline also EMAILS the merchant that reason through the workspace's own decline template once the close has committed; send notifyMerchant false for silence, or notifyTemplateId to pick the words. The send can never fail the close — no address on file, an opted-out contact or a refused rail all come back in notSentBecause beside a closing that landed. Use it instead of a bare status change whenever the deal is actually over. POST /v1/definitions/:key/rows/:rowId/close (access: def:read)

post_definitions_key_rows_rowId_condition-shares

ChatGPT
Share a deal's checklist with its lender Mints a read-only link to the deal's Lender Conditions for the lender and answers {id, url, expiresAt} — the url carries the token ONCE, so a link not saved now is gone. The page shows each condition's name, requirement, who owes it, status, due date, when its last file arrived and how many files it has: never the files, notes, people or other deals. Optional submissionShareId ties the link to a submission of this deal (withdrawing it ends the link); optional expiresAt (RFC3339, default 30 days, at most 90). Needs a seat that writes the deal. Deals page only. POST /v1/definitions/:key/rows/:rowId/condition-shares (access: def:read)

post_definitions_key_rows_rowId_contract_documents

ChatGPT
Add a document to a deal's contract Renders one document against a deal and stores it, pinned to the template version it was built from. kind names the document (or templateId a specific template), and terms supplies values the template asks for beyond the record's own cells — placeholders are filled from the row's cells keyed by slugged column label, plus the workspace name and the caller's email. A placeholder nothing filled is a 422 that NAMES the missing ones rather than issuing a paper with a hole in it; an unknown kind is a 400 and a stranger's template a 404. POST /v1/definitions/:key/rows/:rowId/contract/documents (access: def:read)

post_definitions_key_rows_rowId_contract_documents_preview

ChatGPT
Preview a document on a record Fills a template — a published version or draft by templateId, the current one of a kind, or unpublished title/body words — against one record and answers it WITHOUT storing anything: the filled title and body (an unfilled {{name}} is left in place), values (every variable the words use with its value on this row), missing (names nothing filled — a render would refuse them) and unknown (names neither this board's columns nor the global vocabulary know — a publish against this board would refuse them). Reach for it before issuing a paper, to see what the row lacks and to propose terms for review. POST /v1/definitions/:key/rows/:rowId/contract/documents/preview (access: def:read)

post_definitions_key_rows_rowId_contract_send

ChatGPT
Send the contract for signature Mints one signing link per listed signer, freezes the paper being signed, moves the contract to pending_signature and emails each signer. One transaction. POST /v1/definitions/:key/rows/:rowId/contract/send (access: def:read)

post_definitions_key_rows_rowId_convert

ChatGPT
Convert a lead into a deal Turns one lead into a deal, ONCE and atomically: the mapped cells are copied, the merchant is found or made and linked both ways, the lead's record group and owners are carried onto the deal, the lead is stamped converted and both records get their timeline entries. Answers the new deal row, which is what the app opens. A second call on the same lead is a 409 naming the deal it already made. Leads page only. Many at once: post_definitions_key_rows_act POST /v1/definitions/:key/rows/:rowId/convert (access: def:read)

post_definitions_key_rows_rowId_credit-records_family

ChatGPT
Record a credit record Record one credit record on this deal: a credit memo, a DECISION (approve, decline, counter) or a piece of business research. An approval names an amount, a decline names a reason and a counter names its counter - each enforced by the schema, not by a validator one path can skip. A decision that is not pending rings the deal's people. POST /v1/definitions/:key/rows/:rowId/credit-records/:family (access: def:read)

post_definitions_key_rows_rowId_debt-schedules

ChatGPT
Add a debt schedule to a record Start a business debt schedule on a record. The body carries the whole content plus the revision it is written against; a stale revision is refused rather than overwriting somebody else's edit. POST /v1/definitions/:key/rows/:rowId/debt-schedules (access: def:read)

post_definitions_key_rows_rowId_debt-schedules_scheduleId_issue

ChatGPT
Finalize a business debt schedule revision Issue a business debt schedule: the revision named in the body becomes the one that stands, so it can be sent out and quoted against. Use it when the schedule is final, not while it is being filled in. POST /v1/definitions/:key/rows/:rowId/debt-schedules/:scheduleId/issue (access: def:read)

post_definitions_key_rows_rowId_disclosure

ChatGPT
Generate a submission's financing disclosure Renders the state-mandated commercial financing disclosure for one submission record from an offer's figures and files it. Name the merchant's two-letter state and the offer row, plus any term no offer cell holds — prepayment terms above all. A live disclosure already on the submission is voided and replaced. 422 names any figure the paper must state and could not fill; 400 when the state is not one this workspace discloses in. POST /v1/definitions/:key/rows/:rowId/disclosure (access: def:read)

post_definitions_key_rows_rowId_document-requests

ChatGPT
Record an outstanding document requirement Records that a document is owed on one record: documentType (required), an optional note and an optional RFC 3339 dueAt. A second OPEN ask for the same document on the same record is a 409 that NAMES the ask already standing, so the caller can chase or withdraw that one. IT SENDS NOTHING — this writes down the need; asking the merchant for it is the stipulations route. POST /v1/definitions/:key/rows/:rowId/document-requests (access: def:read)

post_definitions_key_rows_rowId_files

ChatGPT
Attach a file to a record Declares a file on one record and answers the file row plus a signed PUT uploadUrl the client sends the bytes to — this route never carries the file itself. name is required and sizeBytes must not be negative; contentType, documentType and periodStart/periodEnd type the file so the statements and processing reads can find it. Refused when file storage is not configured for this workspace. POST /v1/definitions/:key/rows/:rowId/files (access: def:read)

post_definitions_key_rows_rowId_files_attached

ChatGPT
Upload an attached file to a record Lands a PDF the person attached to their Neo message on one record — a deal, a lead, any board's row — bytes and all, in one step: the file is filed under the caller's seat, stored and landed, and landing starts the record's document processing within a minute (post_definitions_key_rows_rowId_process starts it at once). Send attachment with the file's name exactly as the message lists it, and documentType when you know what the document is; the file's bytes are supplied at the door — fileData and name are never sent by you. This is the upload for a file in the chat: post_definitions_key_rows_rowId_files only declares one and answers an upload URL. A lender's or bank's needs list goes onto the deal this way, and processing creates the deal's conditions and their tasks — never create those by hand beside it. Callable only from a Neo turn that carries the file; refused when file storage is not configured. POST /v1/definitions/:key/rows/:rowId/files/attached (access: def:read)

post_definitions_key_rows_rowId_folders

ChatGPT
Create a document folder Make an empty folder on a record. Omit parentFolderId (or send null) to put it at the record's Documents root; pass another folder's id to nest it. A name already used at the same level of the same record is a conflict. POST /v1/definitions/:key/rows/:rowId/folders (access: def:read)

post_definitions_key_rows_rowId_form-link

ChatGPT
Send an application form link by email or SMS Text or email one record its application form and file the thread under that record. Name a formId (one of the workspace's own published forms) or paste a url, never both. Add prefill:true — formId only — to send the form already filled in from this record and bound to it, so what the merchant sends merges back into the same row instead of opening a second one. Medium is sms or email; the recipient is read off the record when none is given. Reach for it when a desk wants a merchant to complete an application. POST /v1/definitions/:key/rows/:rowId/form-link (access: def:read)

post_definitions_key_rows_rowId_intake-review_changeId_accept

ChatGPT
Accept an intake conflict Settle one open intake item on the record. A held value is written to its column as the caller, through the row's own door, so the board's write level and the column's lock apply; a duplicate mark is recorded as confirmed and nothing is written or merged. An item already settled answers 409. POST /v1/definitions/:key/rows/:rowId/intake-review/:changeId/accept (access: def:read)

post_definitions_key_rows_rowId_intake-review_changeId_reject

ChatGPT
Reject an intake conflict Settle one open intake item on the record without writing anything: a held value stays off the record, a duplicate mark is recorded as not a duplicate. An item already settled answers 409. POST /v1/definitions/:key/rows/:rowId/intake-review/:changeId/reject (access: def:read)

post_definitions_key_rows_rowId_lender-form-fills_documentId_print

ChatGPT
Print a fresh copy of a form issued on a record Draws the form issued on this record (documentId) with every value as it stands now — corrections included, never a signature — files it on the record as the next numbered copy (earlier copies are kept, never overwritten) and answers it with its PDF (base64), for another signature in ink. Requires write access to the record. POST /v1/definitions/:key/rows/:rowId/lender-form-fills/:documentId/print (access: def:read)

post_definitions_key_rows_rowId_lender-forms

ChatGPT
Upload a lender form Turn a PDF already uploaded to this row (fileId) into a lender form filled from definitionKey's records — one of the list's filledFrom pages or this record's own page; left out, the only filledFrom page when there is exactly one. Its blanks are read from its own fillable fields, else from processor-v2's layout of the page (underscores, drawn lines, boxes, checkboxes, signature lines), and each is mapped to the page's words with a confidence; signature, initial and date lines are given a signing party. A bank PDF any workspace already uploaded (the same bytes, or a re-saved copy of the same fillable form) is recognized from the shared form library — no layout is read again — and its blanks are mapped from the concepts its labels and other teams' confirmed mappings name onto THIS workspace's words; a new one has its blanks labeled once for every workspace (Jev, then gpt-6.1-sol, then gpt-6-astra, from the blank form's printed words only). Answers version 1 with the words, suggestions and a note saying how the blanks were found and labeled. Requires write access to the row. POST /v1/definitions/:key/rows/:rowId/lender-forms (access: def:read)

post_definitions_key_rows_rowId_lender-forms_formId_issue

ChatGPT
Issue a lender form for signature Fill a lender form from this record, file the filled PDF on the record, and issue it as a paper (kind lender_form). Send signers: the address each party on the form signs as; sender, {fieldId: value}, the blanks the form leaves to the person sending (refused while a required one is empty); and signature, electronic (the default) or wet — the signer fills their blanks on the signing page, then downloads the form, prints it and signs it in ink. Works the same for a lender's form and a workspace's document template form. Then send it for signature with post_documents_id_signing naming the same people; when the last one signs electronically, every mark is drawn at its place on the form's page and the executed PDF is filed on the record. Sends nothing itself. POST /v1/definitions/:key/rows/:rowId/lender-forms/:formId/issue (access: def:read)

post_definitions_key_rows_rowId_lender-forms_formId_preview

ChatGPT
Preview a lender form filled from a record Fill a lender form from this record without writing anything: values (each field and what it is filled with), missing (mapped fields the record holds nothing for), signatures (the places each party signs), sender (the blanks the person sending fills, each with its value so far) and senderEmpty (the required ones still empty), and pdf, the filled form as base64. An optional body {sender: {fieldId: value}} draws the sender's answers so far. POST /v1/definitions/:key/rows/:rowId/lender-forms/:formId/preview (access: def:read)

post_definitions_key_rows_rowId_merchant-invitations

ChatGPT
Invite a merchant to their portal Mints a merchant-portal invitation for one deal and answers the bearer token ONCE, with the accept path and a full inviteUrl when baseUrl is given — the row keeps only the token's hash, so a token not saved now is gone. email is required, expiresAt must be in the future if sent, and language defaults to en. The merchant company is resolved from the deal's merchant cell at mint time and never re-followed. Deals only. POST /v1/definitions/:key/rows/:rowId/merchant-invitations (access: def:read)

post_definitions_key_rows_rowId_notes

ChatGPT
Add a record note Append a new internal note without replacing earlier notes. Send body and a fresh UUID requestId; reuse that requestId with the same body when retrying a lost response. The server records the signed-in author and creation time. Requires write access to the record. POST /v1/definitions/:key/rows/:rowId/notes (access: def:read)

post_definitions_key_rows_rowId_offers

ChatGPT
Add an offer to a deal Two asks on one path, and the body says which: send {offer: …} and the ONE offer the desk wrote is created and presented in the same transaction; send a riskProfile or a score (0-100, one or the other, never both) and the tier ladder is generated from it. Either way the answer is the offers rows this call created, so a client reads one shape. An authored offer needs a positive funded amount, and a workspace that deleted its presented status has deleted the authored path with it. Deals page only. POST /v1/definitions/:key/rows/:rowId/offers (access: def:read)

post_definitions_key_rows_rowId_plaid_exchange

ChatGPT
Finish a bank link for a deal Trades the public token Plaid Link handed the browser for a long-lived access token, seals it, reads the item's accounts, and records who consented from where. Re-running it for a bank already linked repairs that link rather than adding a second one. POST /v1/definitions/:key/rows/:rowId/plaid/exchange (access: def:read)

post_definitions_key_rows_rowId_plaid_items_itemId_check-payment

ChatGPT
Check an expected bank payment Compare posted transactions against dynamic expectedAmountMinor, dueAt (RFC3339), graceHours, currency, query and direction. Returns matched, not_due, not_observed or unknown. A not_observed result is a review signal, not proof of nonpayment. Unknown or stale data must not trigger a demand. POST /v1/definitions/:key/rows/:rowId/plaid/items/:itemId/check-payment (access: def:read)

post_definitions_key_rows_rowId_plaid_items_itemId_refresh

ChatGPT
Request fresh bank transactions Request an institution refresh for a connected bank. Returns 202 refreshRequested, not completed data. Changes arrive through the Plaid webhook and sync. Five-minute per-bank cooldown returns 429. Requires Plaid Transactions Refresh enabled. POST /v1/definitions/:key/rows/:rowId/plaid/items/:itemId/refresh (access: def:read)

post_definitions_key_rows_rowId_plaid_items_itemId_sync

ChatGPT
Sync one merchant bank Import available transaction updates for this merchant bank only. Use trigger.rowId and trigger.itemId in an enrolled_merchant_banks monthly schedule. A hasMore response is incomplete: do not treat it as a complete review. POST /v1/definitions/:key/rows/:rowId/plaid/items/:itemId/sync (access: def:read)

post_definitions_key_rows_rowId_plaid_items_itemId_transactions_search

ChatGPT
Search imported bank transactions Read bank transactions with currency, date range or lookbackDays, accountId, query and credit/debit direction. Signed minor units: credits positive. Matched count and total cover all matching rows; transactions are capped at 500 and truncated is explicit. lastSyncedAt names the last complete import, not a guarantee the institution is current. POST /v1/definitions/:key/rows/:rowId/plaid/items/:itemId/transactions/search (access: def:read)

post_definitions_key_rows_rowId_plaid_sync

ChatGPT
Pull a deal's linked banks Fetches new, changed and removed transactions for every bank linked to this deal and rewrites the monthly statement periods and positions computed from them. Idempotent - a second run over the same month recomputes the same figures. Answers how many lines arrived and how many months now stand; hasMore true means press it again. POST /v1/definitions/:key/rows/:rowId/plaid/sync (access: def:read)

post_definitions_key_rows_rowId_prefill-links

ChatGPT
Mint a prefilled form link to copy Mint a link to formId prefilled from THIS record and answer it — url and embedHtml — WITHOUT sending anything, for a rep to paste into their own email or chat. The same token post_definitions_key_rows_rowId_form_link mints with prefill:true: what the merchant sends merges back into this record (no second record), the rep who minted it becomes the record's owner if it has none, and files attached on the link land on the record. The form must be published and must file onto this record's own page. The url comes back once; the row keeps only its hash. POST /v1/definitions/:key/rows/:rowId/prefill-links (access: def:read)

post_definitions_key_rows_rowId_process

ChatGPT
Process a record's documents Sends one record's documents to the processor and answers the job. Three answers, and the middle one is the point: 201 when a job was created, 200 when the idempotencyKey already named a job — the reload case, where the caller is told THAT job rather than an error — and 409 would_supersede when the read would overwrite fields somebody entered by hand, with those fields NAMED so the caller can retry with replace: true. A record with no documents on it is refused: there would be nothing to read. POST /v1/definitions/:key/rows/:rowId/process (access: def:read)

post_definitions_key_rows_rowId_processing-feedback

ChatGPT
Record feedback on a processed financial metric Record feedback on a processor-derived financial metric using the exact report, processing job and period identity. Negative feedback is retained with its server-derived value snapshot for authorized System Ops review. POST /v1/definitions/:key/rows/:rowId/processing-feedback (access: def:read)

post_definitions_key_rows_rowId_reactivate

ChatGPT
Reactivate an archived record Puts an archived record back on the live board, in the position it was archived from, with nothing about it changed. Find the row to name here with GET .../rows?archived=true. A record that is already live is not found. POST /v1/definitions/:key/rows/:rowId/reactivate (access: def:read)

post_definitions_key_rows_rowId_script_assist

ChatGPT
Ask for help with a record's call script What to say next: send what the prospect said, and the lead's own script answers with the branch that matches — the line to say, its trigger, its tag and where it goes next. It never writes, and it reads THIS workspace's script, so a workspace that wrote its own objection branches gets its own lines back. Always 200 when the body is well formed: 'nothing in this script covers that' is an answer, carried as matched: false. No model is asked. Leads page only. POST /v1/definitions/:key/rows/:rowId/script/assist (access: def:read)

post_definitions_key_rows_rowId_script_generate

ChatGPT
Generate a call script for a record Builds a call script from one lead's own facts AND SAVES it as that lead's own version — which is what makes the Generate button mean something. Regenerating retires the previous row-scoped version in the same transaction, so the lead's ladder never forks. The answer carries ai_status: disabled and the script is stored as static: no model is wired here, and a static script must never be mistaken for an AI one. Leads page only. POST /v1/definitions/:key/rows/:rowId/script/generate (access: def:read)

post_definitions_key_rows_rowId_seat_seat

ChatGPT
Seat one team on a record Puts one of the workspace's own non-account records into a named seat on a row, or clears it — today a deal's TEAM, from the teams roster. The team's name is resolved here and stored beside its id, so every surface that only reads the cell shows the name; a team from another workspace is refused. People are not seated this way: use the row's assignments door for those. POST /v1/definitions/:key/rows/:rowId/seat/:seat (access: def:read)

post_definitions_key_rows_rowId_snooze

ChatGPT
Snooze a lead Takes one lead out of the due list until a date and writes that date onto its follow-up cell. Body {"until": RFC3339}; a null until un-snoozes and leaves the follow-up date where it is. Refuses a date in the past. Reach for it when somebody says 'not until Tuesday' - editing the follow-up cell alone changes the intention and does NOT hide the row. POST /v1/definitions/:key/rows/:rowId/snooze (access: def:read)

post_definitions_key_rows_rowId_status

ChatGPT
Change a record's status Moves one record to a status its page carries, with optional notes, and writes the move to the record's timeline in the same transaction — which is why this exists rather than patching the status cell. Re-picking the status the record already has is a no-op, not a refusal. status is a status's value or its label, any case ("Warm Lead" moves to warm_lead). A word the page does not carry (refused naming the statuses there are), a word that has been archived, a page with no status column, and (on leads) a move the lifecycle forbids are each refused with a sentence saying so. Money boards refuse: their rows move by their own ledger action. Many at once: post_definitions_key_rows_act POST /v1/definitions/:key/rows/:rowId/status (access: def:read)

post_definitions_key_rows_rowId_submissions_mailer-actions

ChatGPT
Queue a lender submission email action Queues emailout, followup, reply or request_contract using server-resolved recipients, documents and authorized sender. Requires submission write permission and validates the action's prerequisites. Initial sends require a native prepared share and use one canonical operation key. Returns an asynchronous job, not a delivery confirmation. POST /v1/definitions/:key/rows/:rowId/submissions/mailer-actions (access: def:read)

post_definitions_key_rows_rowId_submissions_mailer-actions_jobId_reconcile

ChatGPT
Reconcile an uncertain lender mailer operation Rechecks a failed or uncertain operation using its existing durable operation ID and immutable payload. Requires submission write permission. Reconciliation only polls existing mailer history and never starts a new send when that history is missing. POST /v1/definitions/:key/rows/:rowId/submissions/mailer-actions/:jobId/reconcile (access: def:read)

post_definitions_key_rows_rowId_submissions_prepare-send

ChatGPT
Submit one deal to many lenders Writes one submission record and one cross-tenant share per named lender in a single transaction, then queues the workspace submission template with reviewed deal notes and documents by name. The subject and body come from the standard template; each share preserves its prepared message. Answers a line per lender: the records made, or why that lender was refused. POST /v1/definitions/:key/rows/:rowId/submissions/prepare-send (access: def:read)

post_definitions_key_rows_rowId_subrecords

ChatGPT
Attach or create a subrecord Hang a child record under this one. Pass childKey plus childRowId to attach an EXISTING record, or childKey plus cells to create a new child on that page and attach it in one transaction. The child page must be one this page accepts (see GET /v1/definitions/:key/child-pages), and the database refuses an undeclared pair. This is how a record holds MANY items (an order's line items): one child per item, never a select or a list in one cell. POST /v1/definitions/:key/rows/:rowId/subrecords (access: def:read)

post_definitions_key_rows_rowId_super-kyc

ChatGPT
Create a Super KYC link Mints a Super KYC link on this record asking the merchant to connect their mailbox — Google (Gmail and Drive) or Microsoft (Outlook mail), read-only — and/or their bank through Plaid. Body: email, optional name, sources [google, microsoft, plaid], baseUrl for the link. The token and link are answered once. Once the merchant agrees and connects, Specter keeps watching for other funders, agreements, statements that disagree with the bank and whether the account holder is the business. POST /v1/definitions/:key/rows/:rowId/super-kyc (access: def:read)

post_definitions_key_rows_rowId_super-kyc_id_flags_flagId_acknowledge

ChatGPT
Acknowledge a Super KYC finding Marks one finding seen. It stays in the link's history and leaves the open count. POST /v1/definitions/:key/rows/:rowId/super-kyc/:id/flags/:flagId/acknowledge (access: def:read)

post_definitions_key_rows_rowId_threads

ChatGPT
Start a discussion thread on a record Opens an internal thread on one record. subject is optional — a thread opened from a record needs no title — and participants is optional too, since the opener is always seated on their own thread. Several threads per record is deliberate: one deal legitimately carries 'stips before we submit' and 'why did Lender B decline' at once. Answers the new thread, whose id addresses its messages and its roster. POST /v1/definitions/:key/rows/:rowId/threads (access: def:read)

post_definitions_key_rows_rowId_transition

ChatGPT
Move a deal's status Move one deal to another status. The STAGE is derived from that status's band and written, never taken from the caller; the stage-entry gate is checked on a crossing and may be skipped with force. Send cells to fill what the gate needs in the same transaction, and reason to say why - a terminal status asks for one. For a deal that is actually over, use the close route instead: it records the outcome, the reason and the follow-up task with the move. POST /v1/definitions/:key/rows/:rowId/transition (access: def:read)

post_definitions_key_rows_rowId_underwriting

ChatGPT
Create a deal underwriting file and initial scorecard Opens a deal's underwriting file and lays out version 1 of the scorecard: a pending answer for every ACTIVE criterion in the workspace's template, so the tab draws the whole checklist rather than growing it as people answer. ONE per deal — a second start is a 409. Deals page only. POST /v1/definitions/:key/rows/:rowId/underwriting (access: def:read)

post_definitions_key_scripts

ChatGPT
Add a call script to a page Writes version 1 of a call script for one target on the LEADS page. scope is workspace, status (with statusValue and no row) or row (with rowId and no status); name is required and spine must carry at least one section, while brief and branches hold the opening notes and the objection handling. Workspace admin — a workspace- or status-scoped script is words every rep on the page will read; the door a REP uses for one lead is post_definitions_key_rows_rowId_script_generate. POST /v1/definitions/:key/scripts (access: def:admin)

post_definitions_key_statuses

ChatGPT
Add a status to a page Adds a status to a page's vocabulary: value is the stored word and must be snake_case, label is what a person reads (up to 60 characters), tone is one of secondary, brandLight, brand, warning, success, destructive or a lowercase #rrggbb (default secondary), groupLabel files it under a section and isTerminal marks it an end state. The new status goes last in the order; a value the page already carries is a 409. Workspace admin — the vocabulary is shape, while moving a record between statuses is any member's. POST /v1/definitions/:key/statuses (access: def:admin)

post_definitions_key_threads_threadId_messages

ChatGPT
Post a message on a thread Posts a message to a thread. Anyone @-named in mentions or in the body who is not already on the thread is SEATED on it, because naming somebody in a conversation they cannot see is a mention that does nothing; the named get a mention notification and the rest of the roster a message one. A post to a RESOLVED thread reopens it — somebody having more to say is the definition of not settled. A blank body is refused. POST /v1/definitions/:key/threads/:threadId/messages (access: def:read)

post_definitions_key_threads_threadId_messages_messageId_pin

ChatGPT
Pin a thread message Pins the message a thread keeps at the top, or with {pinned: false} unpins it. Anyone on the thread may pin, not only the author: what matters is the thread's business. Answers the message. POST /v1/definitions/:key/threads/:threadId/messages/:messageId/pin (access: def:read)

post_definitions_key_threads_threadId_participants

ChatGPT
Add someone to a thread Seats one colleague on a thread, named by userId. A person who holds no seat in this workspace is a 404 rather than a silent skip — this call asserts that somebody belongs here, so answering 200 to a request that did nothing would leave a roster on screen the server never wrote. Adding somebody already on the thread is harmless. Answers their participant row. POST /v1/definitions/:key/threads/:threadId/participants (access: def:read)

post_definitions_key_threads_threadId_read

ChatGPT
Mark a thread read Marks the caller's place in a thread and answers how many messages are STILL unread for them. The count is the answer rather than an empty 200, so a caller can tell 'cleared' from 'you were never on this thread' without a second read. POST /v1/definitions/:key/threads/:threadId/read (access: def:read)

post_definitions_key_threads_threadId_resolve

ChatGPT
Resolve a thread Settles a thread, or with {resolved: false} reopens it — one route for both directions so the word and the stamp cannot disagree. Resolving is a STAMP, not a wall: the thread keeps its transcript, keeps answering reads and keeps taking messages, and a new message reopens it. What it stops is the list counting the thread as live work. POST /v1/definitions/:key/threads/:threadId/resolve (access: def:read)

post_definitions_key_underwriting_id_evaluate

ChatGPT
Evaluate a deal against an underwriting rule set Runs the machine over a deal's scorecard and answers the NEW version's rows. It mints version n+1, COPIES every answer from version n into it (nothing in n is ever touched), adds a pending row for any active criterion the template has gained since, then answers the criteria this service can decide by itself — leaving alone any a person has already decided. Reach for it after new documents land; a person's own answers are saved with put_definitions_key_underwriting_id_criteria instead. Deals page only. POST /v1/definitions/:key/underwriting/:id/evaluate (access: def:read)

post_definitions_key_views

ChatGPT
Save a view of a page Saves a view on a record page: its key (a slug), label (up to 80 characters), and the filter, sort and columnIds it restores. scope defaults to user, which is the caller's own private view; workspace makes it everyone's and needs a workspace admin. A key already used in that scope is a 409. This is the one write under a page's shape that an ordinary member may make, because a user view is theirs. POST /v1/definitions/:key/views (access: def:read)

post_document-requests_id_cancel

ChatGPT
Cancel a document request Withdraws one document request — we are no longer waiting on this paper. No body; id is the request. Only an OPEN request can move, so cancelling one already fulfilled or cancelled is a not-found. It drops off get_document-requests afterwards. Reach for post_document-requests_id_fulfil instead when the file actually arrived. POST /v1/document-requests/:id/cancel (access: workspace)

post_document-requests_id_fulfil

ChatGPT
Fulfil a document request Closes one document request because the paper ARRIVED. Body: fileId, required — the file must already be pinned to the same record the ask was raised on, and another workspace's file cannot be named. Only an OPEN request can move, so fulfilling twice is a not-found. Answers the settled request. Upload the file first, then name it here; post_document-requests_id_cancel is for the ask that is simply dropped. POST /v1/document-requests/:id/fulfil (access: workspace)

post_document-templates

ChatGPT
Create a document template Publishes the next version of one kind of paper this workspace issues. Body: kind (fee_agreement or financing_disclosure), title and body are the words; state is the two-letter US state they are written for and absent means every state, which is what a fee agreement always is; bodyFormat is markdown (the default) or text. There is no edit and no delete by design — a version's words are immutable, so correcting them is publishing version n+1, and paper already issued keeps pointing at the version it was made from. A workspace admin's, because the words the company issues bind the company. POST /v1/document-templates (access: workspace:admin)

post_document-templates_forms

ChatGPT
Upload a document template form Turn a PDF already uploaded among the workspace's own files (fileId) into a document template form filled from definitionKey's records — any board: a law case, a deal, a CRE loan, a made board. Its blanks are read from its own fillable fields, else from processor-v2's layout, recognized from the shared form library when any workspace uploaded the same form, and each is mapped to the page's words; signature, initial and date lines are given a signing party. Answers version 1 with the words, suggestions and a note saying how the blanks were found. Admin only. POST /v1/document-templates/forms (access: workspace:admin)

post_document-templates_suggest-words

ChatGPT
Place a template's words in an uploaded document Reads a document a firm uploaded (text, as extracted) against the words a template written for definitionKey may use — the board's own and every linked board's (get_definitions_key_template-words) plus the workspace's (get_message-placeholders) — and answers where each field sits, writing nothing: spans (each {start, end, text, word}, UTF-16 offsets into the text sent, word always one of those words), maybe (runs that read like a blank to fill but that no word means, each with why), invented (words the model answered that are not in the list, dropped), and body (the text with every span replaced by its {{word}} and no other character changed — an answer that would change the wording is refused with 422). The text is sent to the model for this call only and is not stored or logged. Publish the reviewed words with post_document-templates. A workspace admin's, the publish's own door; 503 when Neo has no model key, 504 when the model does not finish. POST /v1/document-templates/suggest-words (access: workspace:admin)

post_documents_id_letter

ChatGPT
Send a document as a letter Mail one rendered paper as a letter, not for signature: a demand to a carrier, a letter of representation to an adjuster. Name the addressee (to.name, to.email); optionally from, an email line you may send from, and message, a covering note. The paper is drawn as a PDF, filed on its record under the paper's own file type, and mailed on the record's thread with that PDF attached, from the workspace's paper sender (Settings › Email sending) under the workspace's name unless from names a line. Refused with 400 when the workspace has no address to send from, and 503 when the deployment has no file storage; nothing is written either way. POST /v1/documents/:id/letter (access: workspace)

post_documents_id_signing

ChatGPT
Send a document for signature Send one rendered document out for signature: name the signers (name, email and role), optionally an order, an expiry and a covering sentence, and each is mailed a single-use link to a public signing page. Reach for this to put an offer, a funding agreement or a fee agreement in somebody's hands. It supersedes any request still standing on the same paper. Set sequential to make each signer wait for the one before, and includeWorkspaceSigner to seat the workspace's own execution profile last. POST /v1/documents/:id/signing (access: workspace)

post_files

ChatGPT
Upload a file Declare a workspace-loose file and get back its row plus a signed upload URL good for 15 minutes — the bytes go straight from the client to storage and never through this service. Needs a name and a non-negative sizeBytes; documentType, periodStart and periodEnd are optional and can be set afterwards with PATCH /v1/files/:id. Answers 503 when this deployment has no file storage configured, rather than a URL that would not work. POST /v1/files (access: workspace)

post_files_id_analyze

ChatGPT
Analyse a document Send one document to the analysis engine, or re-run it: records a queued analysis at the next attempt number and queues the hand-off. Reach for it when a single document was re-uploaded or came back unreadable; processing every document on a deal is the record-level process route instead. Refuses with 409 while an analysis of that document is already queued or running. POST /v1/files/:id/analyze (access: workspace)

post_files_id_landed

ChatGPT
Confirm a file's upload landed Say the PUT to a file's signed upload URL finished. The server looks in storage before it believes it, and only then is the file real: its record is told, the ask it answers is settled and processing may read it — none of that happens at the declare. Answers 200 with the file and its landedAt; a second confirm answers the same. 409 bytes_missing when storage holds nothing under it yet, 409 size_mismatch with expected and got when it holds a different number of bytes than declared. Reach for it right after uploading the bytes of POST /v1/files, a record's files route or a replace. A file nobody confirms is found by a sweep within minutes, and given up on after a day. POST /v1/files/:id/landed (access: workspace)

post_files_id_links

ChatGPT
Link a file to a record Show one file on one more record, beside the record it is filed on: send the record's page as definitionKey (its key, like "deals" or a board's "board-…") and the record as rowId. Answers the file with its links. The caller must be able to open the file and edit that record (on a board, the write level); a record they cannot reach is a 404. Linking the record the file is already filed on, or one already linked, changes nothing. The file's bytes are not copied — every link shows the same file. POST /v1/files/:id/links (access: workspace)

post_files_id_replace

ChatGPT
Replace a file's contents Put new bytes under an existing file and keep its id, so every record it is on, its classification, folder and requests stay as they are. Send name, contentType and sizeBytes as for a new upload; the answer is the same — the file and a signed upload URL (15 minutes) the client PUTs the new bytes to. The bytes it replaced are kept, and until the new ones land in storage a download still serves the old ones. Refused with 409 while a ported file's bytes are still being moved in. POST /v1/files/:id/replace (access: workspace)

post_files_id_result-changes_changeId_accept

ChatGPT
Accept a document's held change Writes a change held for review into its column, as the caller — their board access and the column's lock apply — and marks it accepted. 409 when the change was already settled. POST /v1/files/:id/result-changes/:changeId/accept (access: workspace)

post_files_id_result-changes_changeId_reject

ChatGPT
Reject a document's held change Keeps a change held for review off the record and marks it rejected; nothing on the record is written. 409 when the change was already settled. POST /v1/files/:id/result-changes/:changeId/reject (access: workspace)

post_files_id_share

ChatGPT
Share a file by link Mint a signed download URL for one file that works for 7 days, for handing it to somebody outside the workspace who has no seat, and answer {url, expiresAt}. Anyone holding the URL can download the file until it expires, and it cannot be revoked early, so reach for the 15-minute download route instead when the person asking will open it themselves. The share is filed on the workspace's activity feed. Refused with 409 while a document is locked waiting for its unlocked canonical PDF, and 503 when this deployment has no file storage. POST /v1/files/:id/share (access: workspace)

post_forms

ChatGPT
Build a form Creates a form, and always as a DRAFT — post_forms_id_publish is what opens the public door afterwards. Body: name, an optional slug (lowercase letters, numbers and hyphens; the words drafts and submissions are refused because the public router already means something by them), fields, sections, settings, and targetDefinitionKey naming the record page answers should land on. A form that fills a board is post_forms_from-board: it writes the questions from the board's columns and publishes in one call. POST /v1/forms (access: workspace)

post_forms_from-board

ChatGPT
Make a form for a board Builds a form from a board's own columns and PUBLISHES it at once — the one call for 'make a form for this board'. Body: targetDefinitionKey (the board's key; the board on screen when the person says 'this board'), optional name (default: the board's name), description, columnIds (a subset, in the order to ask them; omit for every column a person can fill, in the board's order) and sections ([{title, columnIds}], every chosen column in exactly one; omit for one unsectioned form). The server writes the questions: one per column, the question kind matched to the column kind, starred columns required, each answer landing in its column. Person, relation, file, formula, lookup and files columns are never asked. Prefer this over 'Build a form' whenever the form fills a board — never hand-write the field list for one. Answers the live form with its slug; 'Where this form lives on the web' gives its address. POST /v1/forms/from-board (access: workspace:admin)

post_forms_id_duplicate

ChatGPT
Duplicate a form Copies ONE form into a NEW draft with its own id and its own web address: every question (ids, rules, board columns), every page, the board it fills and the whole settings document (look, logo, pacing, analytics, saved drafts, application PDFs, deal matching, after-submit, edit links, agreements, automations, emails) come across; its responses, visits, counters, old addresses and people's links stay with the original, which is not touched. Body: optional name (default: the original's name plus " (copy)", or " - <person>" when made for someone) and optional forUserId, an active member's account id: the copy then carries that person's own link, so what comes in through it is theirs — admin only, like giving a link. Answers the copy, repLinkId when one was made, and notes, a sentence for anything that could not come across as it was (an email inbox the caller may not send from). Publish the copy with 'Publish a form'. POST /v1/forms/:id/duplicate (access: workspace)

post_forms_id_publish

ChatGPT
Publish a form Opens the public door on one form: it starts answering at its slug and a stranger holding the link can file it. Workspace admin, because publishing opens a door strangers walk through. POST /v1/forms/:id/publish (access: workspace:admin)

post_forms_id_rep-links

ChatGPT
Give people their own link to a form Mint a per-person link to ONE form for each account in userIds (active, unpaused members of this workspace only; a paused or foreign account refuses the whole call by name). Somebody who already holds a live link gets the same one back, so calling twice never makes two. Answers every link with its url and embedHtml. Admin, like publishing. POST /v1/forms/:id/rep-links (access: workspace:admin)

post_forms_id_submissions_sid_edit-link

ChatGPT
Mint a correction link for one submission Mints a link that lets whoever filed ONE submission correct it, and answers the url ONCE — the row holds only the hash, so a lost link is re-minted rather than looked up. Reach for it when a merchant says they sent a figure wrong. POST /v1/forms/:id/submissions/:sid/edit-link (access: workspace)

post_forms_id_unpublish

ChatGPT
Unpublish a form Closes the public door on one form. Every answer already collected stays exactly where it is, and publishing again reopens the same address — this is what to reach for when a form is finished, rather than delete_forms_id, which takes the answers with it. Workspace admin. POST /v1/forms/:id/unpublish (access: workspace:admin)

post_integrations_managed-zoom_orders

ChatGPT
Create a managed phone order Workspace admins only; deployment must enable managed phone ordering. Requires the user's explicit acceptance of exact verified monthly prices and quantities, pricingAccepted=true, a prepared request, current offer revision IDs and a stable idempotencyKey. Reuse the same key for retries; never change its selections. Creates an awaiting-checkout order, not a purchase, payment or activation. Taxes and usage are separate. Additional numbers alone require an existing active licensed order. POST /v1/integrations/managed-zoom/orders (access: workspace:admin)

post_integrations_managed-zoom_orders_id_billing_inspect

ChatGPT
Check current managed phone subscription Workspace admins only. Read the current Stripe subscription bound to this paid managed phone order and compare customer, prices, quantities, currency, current period, cancellation and retained deletion evidence. Available when new purchases are disabled. Performs provider reads only; it never charges, provisions or activates service. A verified result is a point-in-time subscription match, not an activation grant or proof of paid renewal invoices. POST /v1/integrations/managed-zoom/orders/:id/billing/inspect (access: workspace:admin)

post_integrations_managed-zoom_orders_id_checkout

ChatGPT
Open managed phone payment checkout Workspace admins only; disabled by default. Requires explicit user instruction to open checkout for the selected order and openCheckout=true. Reuses the workspace's billing customer, or creates it using a persisted identity and idempotency key before checkout. Uses frozen order parameters. Can create a Stripe checkout session and subscription when the user pays. Never call without purchase intent. Retries reuse the original session/request; completion still requires payment reconciliation and provider provisioning before activation. POST /v1/integrations/managed-zoom/orders/:id/checkout (access: workspace:admin)

post_integrations_managed-zoom_review

ChatGPT
Request managed phone review Freeze a complete saved draft revision and request internal Specter review. Get the user's explicit request to submit their draft before calling. Repeated calls with the same revision return the same request. This does not file with Zoom, buy a line, authorize billing or claim carrier approval. POST /v1/integrations/managed-zoom/review (access: workspace:admin)

post_integrations_managed-zoom_review_refresh

ChatGPT
Refresh customer campaign approval Refresh Zoom evidence for this workspace's latest review and previously operator-bound campaign. Caller supplies no provider IDs; cannot create or move a campaign binding. Reuses observations younger than one minute. Returns the review with dated provider status or null if no review. No binding means no provider request. Unavailable provider reads must not be interpreted as approval. Does not assign numbers, charge, submit, or authorize sending. POST /v1/integrations/managed-zoom/review/refresh (access: workspace:admin)

post_integrations_managed-zoom_review_reply

ChatGPT
Reply to managed phone campaign review Workspace owner/admin only. Read the current review and ask for the user's actual feedback before calling. Append their response to the selected request's durable history so Specter operators can act on it. Supply requestId from the review, a fresh clientId UUID, and note (1–4000 characters); reuse the exact clientId and note on retries. Never invent consent, business facts, or approval. This only records feedback: it does not change a frozen application, provider approval, billing, phone assignments or SMS permissions. Explain needed next steps and obtain appropriate authorization before separate actions. POST /v1/integrations/managed-zoom/review/reply (access: workspace:admin)

post_integrations_plaid_partner_refresh

ChatGPT
Check the Plaid onboarding with Plaid Ask Plaid where the end customer's review stands and which OAuth banks have registered it, and store the answer. note carries a sentence when the bank list could not be read. Admin only. POST /v1/integrations/plaid/partner/refresh (access: workspace:admin)

post_integrations_provider_test

ChatGPT
Create integrations provider test Open the stored credentials and actually reach the provider. Only twilio and zoom have a probe; every other provider answers the honest 'not checked — the settings are complete, whether they work is unknown' rather than a tick it has not earned. The answer says whether it was checked, whether it was reachable, and why not. A provider with no row is a not-found, because there is nothing to test. Admin only, and the plaintext credential never appears in the response or the log. POST /v1/integrations/:provider/test (access: workspace:admin)

post_integrations_ringcentral_oauth_resume

ChatGPT
Finish connecting RingCentral Finish the caller's consent using the callback query returned to the Specter receiver. Consumes the state bound to this seat, exchanges the code, saves its extension and verifies owned lines. Imports call, recording and incoming/outgoing SMS history; does not enable SMS sending. POST /v1/integrations/ringcentral/oauth/resume (access: workspace)

post_integrations_ringcentral_retry

ChatGPT
Retry RingCentral history sync Retry the caller's RingCentral sync, failed recording downloads and pending transcript enrichment after resolving the connection error. Does not send messages or affect a colleague's connection. Answers 204. POST /v1/integrations/ringcentral/retry (access: workspace)

post_integrations_zoom_oauth_resume

ChatGPT
Finish connecting Zoom Finishes the caller's Zoom consent: query is the callback's query handed back to the app's receiver. Refuses (403) any seat but the one that started it before asking Zoom anything, then exchanges the code, saves the grant to this seat and refreshes its lines. Answers connected: true. Only the desktop's receiver calls it. POST /v1/integrations/zoom/oauth/resume (access: workspace)

post_iso-conflict-checks

ChatGPT
Check a business for ISO conflicts Asks whether this merchant has ALREADY been submitted to this lender, and whether it was somebody else who sent it — the conflict check before a file goes out, so two brokers do not shop the same deal to the same funder. Body: merchantKey and lenderTenantId, the lender's linked workspace from get_lender-workspaces. A POST rather than a GET because the merchant identifier is the question's body, and a merchant's name does not belong in a URL that proxies and access logs keep. The answer names no other broker. POST /v1/iso-conflict-checks (access: workspace)

post_mail-connections_id_operation

ChatGPT
Change a mailbox folder or message flags Create or rename a native Yahoo/AOL folder, move an existing message, or set read/starred/answered/draft flags. Requires your own connected mailbox, write consent and provider activation. Use folder, uid_validity and uid from a message appearance or received event; never guess them. Supply a stable UUID operation_id. An uncertain outcome must be checked in the mailbox before another operation; never retry with a new ID. Operational use only; no advertising. POST /v1/mail-connections/:id/operation (access: workspace)

post_me_screen-capture_allow

ChatGPT
Allow screen capture of my window for a while Lets the signed-in person's own desktop window be screenshot and screen-recorded for a stretch — {minutes}, thirty when the body says nothing, clamped to 1..120. Answers the same shape the read does, mode and style together, and the window is server-side, so it holds on their other machines too. A seat an operator set to locked is refused with 409 screen_capture_locked; one already off is a no-op, because nothing is hidden to permit. One person's own state, and it lands in the workspace's activity feed. POST /v1/me/screen-capture/allow (access: workspace)

post_message-automation-rules

ChatGPT
Save a message automation rule configuration Save a workspace message-rule configuration with its name, channel, record type, trigger event and conditions, template, recipient field, approval preference and enabled flag. This endpoint stores configuration and an audit entry only. The service does not currently execute these rules, schedule messages or send messages when this configuration is saved. To send a message now, use the communication tools; to prepare a message for review, create a message draft. POST /v1/message-automation-rules (access: workspace)

post_message-drafts

ChatGPT
Write a draft for somebody to look over Writes the words of an email or text and queues them for approval instead of sending them: write a draft here when a person should look over the wording before it goes out. When asked to draft an email, use this tool so the email appears in the person's Inbox → Drafts, and tell them where it is saved. Body: recipient (required), body (required), channel (email or sms), subject (email only), optional templateId, and definitionKey plus rowId (both or neither) to preserve the related record. Do not replace saving with a chat-only answer. An email's sender can be chosen in the draft editor later, or remembered with a versioned patch after creating it. A draft request never authorizes sending. Once explicitly asked to send a saved email, use post_message-drafts_id_send with its current updatedAt as expectedUpdatedAt, recipient, subject, body and authorized accountIdentifier. Approval alone sends nothing. POST /v1/message-drafts (access: workspace)

post_message-drafts_id_approve

ChatGPT
Approve a draft Record approval of a draft, stamping who and when. APPROVAL IS NOT A SEND. For a personal saved email, send only when explicitly authorized using post_message-drafts_id_send with the current updatedAt as expectedUpdatedAt, recipient, subject, body and accountIdentifier. That door queues once and returns the same receipt on retry. For a text, post_comms_send dispatches it separately. Cancel bins an unsent draft. POST /v1/message-drafts/:id/approve (access: workspace)

post_message-drafts_id_cancel

ChatGPT
Cancel a draft Closes a queued draft with a reason so nobody acts on it. Body {"reason": "…"}, optional. Cancelling an APPROVED draft is allowed — nothing has gone anywhere — and it clears the approval stamp with it, so the row never claims both. Cancelling an already cancelled one answers not-found rather than stamping it twice. To fix the words instead of binning them, patch_message-drafts_id. POST /v1/message-drafts/:id/cancel (access: workspace)

post_message-templates

ChatGPT
Keep a reusable message body Saves a reusable email or text body with {{placeholders}} in it — the words a workspace uses over and over, filled in per record later. Body: channel ("email" or "sms"), name (required, one per channel), category (one of these, nothing else — email: emailout, followup, request_contract, reply, merchant_decline, term_loan_readiness, facility_status, submission_missing_info, submission_decision, submission_contracts, submission_funding, workflow; sms: follow_up, notification, reminder, welcome, workflow — a first-touch outreach email is emailout), subjectTemplate, greetingTemplate, bodyTemplate (required), closingTemplate, bodyFormat ("text" or "html"), entityKind (which record kind it is written about) and isDefault. A text is body-only and an email must carry a subject line; every placeholder name must be one get_message-placeholders knows. Nothing goes out from here — a template is words waiting to be filled in: post_message-drafts queues a real letter for a person to approve, post_blasts aims the same words at a whole list of people, and post_message-automation-rules has a template used whenever something happens to a record. One letter to one person right now is the comms surface's, not this one. POST /v1/message-templates (access: workspace)

post_message-templates_id_preview

ChatGPT
See what a template would say Fills one template in and answers the rendered subject and body, plus the placeholder names that found no value — so a person sees the real letter instead of a hopeful one. Body: definitionKey and rowId (the record whose facts fill the holes — a lead, a deal), and values for anything typed by hand. All three are optional: a preview with no record answers the template with every placeholder reported missing, which is what a composer shows before a record is picked. Nothing is stored and nothing is sent. POST /v1/message-templates/:id/preview (access: workspace)

post_message-templates_id_set-default

ChatGPT
Make a template the default Makes this template the default for its channel and category, displacing whoever held that slot — the default is what a composer offers first, and the one a route writing its own letter picks up. One default per (channel, category), held by a unique index rather than by hope. Takes no body; the channel and category come off the row. POST /v1/message-templates/:id/set-default (access: workspace)

post_message-templates_seed-defaults

ChatGPT
Give a workspace the starter templates Creates the starter shelf for one channel — one template per category, each claiming its category's default slot only if that slot is empty. ?channel=email|sms is required. Idempotent: running it twice creates nothing the second time. For a workspace whose template list is empty and needs somewhere to start; to write your own instead, post_message-templates. POST /v1/message-templates/seed-defaults (access: workspace)

post_neo_evaluate

ChatGPT
Evaluate workflow data with Neo Wrap a custom prompt and input JSON in a tool-free model request with strict structured output. outputs is a list of {key,type,description}; types: string, number, boolean, strings. Returns result, summary, evidenceIds and uncertain. Evidence IDs must exist in input. Null means unknown; uncertain results should route to internal review. POST /v1/neo/evaluate (access: workspace)

post_neo_memories

ChatGPT
Remember something about you Saves one thing for Neo to remember about the CALLER in every later conversation — "remember that…", "from now on always…", "keep in mind…". kind is rule for how they want Neo to work (the line to send from, a P.S., no amounts, show me first) or fact for something about them or their desk (their territory, the lenders they use); text is one line in their own words, at most 500 characters. Words already saved (case and spacing aside) answer 200 with that memory and add nothing, so saying it twice is safe; a new one answers 201. A person keeps at most 100: the 101st is refused with 409 neo_memories_full, so forget one first. The conversation it was said in is recorded by the server, never sent in the body. POST /v1/neo/memories (access: workspace)

post_notification-types

ChatGPT
Add a notification type Adds one of the workspace's own notification kinds — key, label, description, category, priority, and whether it shows and makes a sound by default — so it can be raised and everybody can set a preference for it. Admin only. POST /v1/notification-types (access: workspace:admin)

post_notification-types_key_raise

ChatGPT
Raise a notification Sends one of the workspace's OWN notification types to the recipients named, with a title and a body. Specter's own types are refused: those mean something the service vouches for, and a hand-raised one would make the feed unciteable. Admin only. POST /v1/notification-types/:key/raise (access: workspace:admin)

post_notifications_dismiss

ChatGPT
Dismiss notifications Takes the notifications named in {ids: […]} off the caller's own feed. POST /v1/notifications/dismiss (access: workspace)

post_notifications_mark-all-read

ChatGPT
Mark every notification read Clears the caller's whole bell in one write. POST /v1/notifications/mark-all-read (access: workspace)

post_notifications_mark-read

ChatGPT
Mark notifications read Marks the notifications named in {ids: […]} as read, for the caller only. POST /v1/notifications/mark-read (access: workspace)

post_obligations

ChatGPT
Raise an obligation Puts a new thing owed on a deal — a stipulation, a document, reporting, a covenant check, insurance, a licence renewal, a review, a repayment. Body: dueAt is required and so is a title; dealRowId (the deal record on our own board) or facilityId says what it is for, plus obligationType (repayment, document_submit, stipulation, reporting, covenant_check, insurance, license_renewal, review), priority (low, medium, high, critical), responsibleParty (borrower, lender, broker, system), description, notes, assignedToUserId, currency, amountMinor (MINOR UNITS), requiredDocumentType, slaHours, and the counterpartyId / submissionId / decisionId / parentObligationId links. With slaHours the breach moment is computed once from the due date, so the pair cannot drift. It opens pending. POST /v1/obligations (access: workspace)

post_obligations_id_complete

ChatGPT
Complete an obligation Marks what was owed as done, stamping who closed it and when. Body: optional notes and a documentId for the file that satisfies it. Legal from anything not already finished. Reach for post_obligations_id_waive when the obligation is being dropped rather than met. POST /v1/obligations/:id/complete (access: workspace)

post_obligations_id_mark-overdue

ChatGPT
Mark an obligation overdue Records that an obligation has blown its due date, stamping the breach severity with it. No body. Legal from pending or due only. It does not chase anybody: the obligation stays owed until it is completed or waived. POST /v1/obligations/:id/mark-overdue (access: workspace)

post_obligations_id_waive

ChatGPT
Waive an obligation Drops an obligation that will not be met — the desk decides it is no longer required. Body: reason is REQUIRED, in the handler and in the table, so a reasonless waiver cannot be written by any writer. Completing one that was actually satisfied is post_obligations_id_complete. POST /v1/obligations/:id/waive (access: workspace)

post_outreach_bulk-send

ChatGPT
Queue personalized messages Queue up to 25 SMS, iMessage or email recipients through the shared send path. Each recipient may override body and subject and carry expected_last_message_id (latest reviewed message, 0 only for verified empty history). Results distinguish queued, skipped and rejected; succeeded means queued, not delivered. Duplicate destinations in one request are skipped. Never blindly replay partial or unknown outcomes. account_identifier is the line the whole batch leaves on, spelled as get_comms_senders lists it; empty is the caller's OWN line when they have one (own: true in get_comms_senders), otherwise the workspace's default line, so name it whenever the person named one for the batch. A line not theirs refuses the whole batch (line_refused) before anything is queued. 25 is a size per call, not a cap on the list: a longer list is more calls. Each recipient's own body and subject make this the send for messages worded per person; the same words to everyone is a blast (post_blasts). A recipient may name its thread with conversation_id (an integer id, or the string get_definitions_key_replies answers as conversationId) instead of to: it is a reply on that conversation, to its address, by its medium (the batch's), on its own line, and an email without subject takes the thread's as "Re: …" — the follow-up to people who have not answered. Put the thread's address in to beside it so the approval names each person; a to that is not the thread's refuses that recipient. A recipient's record ({key, row_id}) files its thread on that record; one without it is filed by its address exactly as post_comms_send files it (a record whose phone or email cell holds it), and each result's records says where it landed. One at a time: post_comms_send POST /v1/outreach/bulk-send (access: workspace)

post_packs_key_install

ChatGPT
Install a pack Installs a pack into the workspace: it gets its own EDITABLE copy of every page in the pack it does not already have, together with the workflows that act on those pages — a rule naming the deals page is meaningless until that page exists, and asking somebody to install a pack and then find a second button is how a feature ships switched off. key is the pack (finance; get_packs lists them and shows what is in them). Idempotent and by key: a page the workspace already has, edited or archived, is left alone rather than restored, so calling it twice is safe. Answers what it added. A workspace admin's. POST /v1/packs/:key/install (access: workspace:admin)

post_packs_key_roles_install

ChatGPT
Install a pack's roles Installs the role presets a pack ships without installing its boards — the onboarding's industry step, or a workspace that installed the pack before it shipped roles. A pack install already brings them. Idempotent and by key: a preset the firm edited is never overwritten. Answers the keys it added. A workspace admin's. POST /v1/packs/:key/roles/install (access: workspace:admin)

post_packs_key_templates_template_install

ChatGPT
Install one board of a pack Installs ONE board of a pack: key is the pack and template the board's template key, as get_packs lists them. The board is the workspace's own editable copy; it is filed into the pack's sidebar folder when the pack names one, and its relation columns link to the pack's other boards already installed — those not yet installed link when they arrive. Seeds no workflows. Idempotent: a board the workspace already has is left alone and added is empty. 404 for a pack or a board the pack does not ship. Answers the pack and what it added. A workspace admin's. POST /v1/packs/:key/templates/:template/install (access: workspace:admin)

post_payment-schedules

ChatGPT
Raise a repayment schedule Creates a repayment schedule on a deal and, when asked, writes its whole instalment run. Body: startDate is required; dealRowId (the deal record on our own board) or facilityId says what it is for, plus scheduleNumber, currency, fundedAmountMinor, totalPaybackMinor, paymentAmountMinor, originationFeeMinor, otherFeesMinor, factorRate, interestRate, totalPayments, paymentFrequency (daily_ach, daily, weekly, bi_weekly, monthly), paymentMethod (ach, wire, check, card, revenue_split, manual, other), offerId, contractId, fundingRecordId, internalNotes, and generatePayments to write the instalments now. Money is MINOR UNITS — dollars in paymentAmountMinor are wrong by a hundredfold. Every instalment is the payment amount and the last absorbs the remainder, so the run sums to the payback exactly; a payment count and amount that overshoot the payback are refused rather than stored. Completing a deal's funding raises a schedule on its own, so reach for this when a schedule is being set up by hand. POST /v1/payment-schedules (access: workspace)

post_payment-schedules_id_complete

ChatGPT
Close out a schedule Marks a schedule finished: it stops collecting, the end date is stamped today and the next payment date is cleared. No body. Legal from active, paused or defaulted — completing an already finished or cancelled schedule means nothing and answers not found. This records that collection is OVER; how the deal ended in money terms — paid in full, an early payoff, a write-off, a settlement — is post_servicing_outcomes. POST /v1/payment-schedules/:id/complete (access: workspace)

post_payment-schedules_id_pause

ChatGPT
Pause collection on a schedule Stops collecting on a schedule and clears its next payment date. Body: an optional reason, kept as the schedule's last modification reason. Legal from active only, so a second press changes nothing and answers not found — which is also what makes the modification count honest. Restart it with post_payment_schedules_id_resume; to close a schedule out for good use post_payment_schedules_id_complete, and to stop a merchant's debits at the source revoke the authorisation with post_servicing_authorizations_id_revoke. POST /v1/payment-schedules/:id/pause (access: workspace)

post_payment-schedules_id_resume

ChatGPT
Resume a paused schedule Restarts collection on a paused schedule. No body. The next payment date comes back from the instalments themselves — the soonest one still scheduled — rather than from a remembered number. Legal from paused only; anything else answers not found. POST /v1/payment-schedules/:id/resume (access: workspace)

post_payment-schedules_id_servicing-authorizations

ChatGPT
Ask the merchant to authorise the debit Raises the merchant's permission to debit this schedule — the signed authorisation a debit needs to be legal. Body: expiresAt is REQUIRED as an RFC3339 time, because an authorisation with no end never lapses; plus contactChannel (email, sms, in_person, phone) and contactTarget for where the request goes, amountMinor (MINOR UNITS, the ceiling per debit), frequency (daily_ach, daily, weekly, bi_weekly, monthly), currency, authorizationText — the exact words the customer is shown, stored so they cannot be edited out from under the consent — sourceReference for the funding source as the provider names it, and the facilityId / counterpartyId links. It opens pending and moves on with post_servicing_authorizations_id_accept. This authorises money; it does not move any. POST /v1/payment-schedules/:id/servicing-authorizations (access: workspace)

post_payments_id_failed

ChatGPT
Mark an instalment failed on the rail Records that a debit was attempted and the rail rejected it, and schedules the retry: the attempt count goes up and the next retry date is set while retries remain. Body: reason is REQUIRED — a failure with no reason is refused — plus an optional code for the rail's own return code. Legal from scheduled, pending or processing. Reach for post_payments_id_missed instead when nothing was ever attempted, and post_payments_id_reverse when a booked receipt has to be undone. POST /v1/payments/:id/failed (access: workspace)

post_payments_id_missed

ChatGPT
Mark an instalment missed Records that an instalment came and went with nothing attempted and nothing collected. Body: an optional reason (or notes). Legal from scheduled or pending only. A missed instalment is still owed — forgive it with post_payments_id_waive; a rail-level rejection is post_payments_id_failed. POST /v1/payments/:id/missed (access: workspace)

post_payments_id_reconcile

ChatGPT
Agree an instalment against the bank Marks a booked receipt as checked against the bank statement, stamping who agreed it and when. Body: an optional notes (or reason). Legal from completed or partial only. The variance is not recomputed — it is a generated column and was already right. get_payment_schedules_id_reconciliation says how many of a schedule's receipts are still unreconciled. POST /v1/payments/:id/reconcile (access: workspace)

post_payments_id_record

ChatGPT
Record money collected on an instalment Books what actually landed on one instalment, with its split, and recomputes the schedule's totals — all in one transaction. Body: actualAmountMinor (MINOR UNITS) is the receipt, plus penaltyAmountMinor, method (ach, wire, check, card, revenue_split, manual, other), externalTransactionId, achTraceNumber, batchId, notes, and allocations as {allocationType, amountMinor, description} over principal, interest, fee, penalty, commission, buyout and other. An allocation list that does not sum to the receipt is refused rather than stored. The instalment lands on completed, or partial when the amount is short. Legal from the unsettled states only, in the WHERE, so two people recording the same receipt cannot both count it. This RECORDS money that already arrived; to actually debit the merchant's bank now, reach for post_payments_id_collect. POST /v1/payments/:id/record (access: workspace)

post_payments_id_reverse

ChatGPT
Reverse a booked receipt Undoes a receipt that was already booked — a bounced or clawed-back payment — and recomputes the schedule's totals so the money stops counting as collected. Body: an optional reason (or notes). Legal from completed or partial only. The amount stays on the row as the historical fact it is. A debit the rail refused in the first place is post_payments_id_failed, not this. POST /v1/payments/:id/reverse (access: workspace)

post_payments_id_waive

ChatGPT
Forgive an instalment Waives one instalment: it stops being owed and stops counting towards the schedule's outstanding balance. Body: an optional reason (or notes), APPENDED to the instalment's notes rather than written over them. Legal from scheduled or missed. Reach for it when the desk decides not to collect; post_payments_id_reverse is for a receipt that was booked and has to be unbooked. POST /v1/payments/:id/waive (access: workspace)

post_reports_preview

ChatGPT
Run an unsaved report definition Answers the numbers for a report definition that has NOT been saved and may never be — the ad-hoc "count these, grouped by that". The body IS the definition, not wrapped in anything: {source, measures, dimensions, filters, limit}, built from get_reports_catalogue. Saves nothing. Use post_reports_views_id_run for a report the workspace already keeps, and post_reports_views to store this definition once it is right. POST /v1/reports/preview (access: workspace)

post_reports_views

ChatGPT
Save a report Saves a new report — a stored QUESTION, never rows, re-read from the records on every run. Body: name, definition ({source, measures, dimensions, filters, limit}, whose source and field keys come from get_reports_catalogue), optional description, and visibility — private (the default) or tenant to share it with the whole workspace. For a one-off number nobody needs kept, post_reports_preview answers without saving anything. POST /v1/reports/views (access: workspace)

post_reports_views_id_run

ChatGPT
Run a saved report THE NUMBERS for a saved report — "the figures for last month", "how many deals funded by stage" — run now, under this caller's own reach, and answered as the view plus a result of columns and rows. A POST because a run does work and nothing caches it: every run re-reads the records, so a revoked seat stops counting immediately. A percent comes back as a fraction 0..1 and a group with no value as null. For a definition that is not saved, post_reports_preview. POST /v1/reports/views/:id/run (access: workspace)

post_reports_views_id_shares

ChatGPT
Publish a public report link Publishes a PUBLIC link to a report: anybody holding the URL can open its numbers with no sign-in, which is why it is an admin's. The token is answered exactly ONCE — only its sha256 is stored, so a lost URL is re-published, never looked up. Body is optional: label and expires_at (RFC3339, and it must be in the future; 30 days by default, since a public link with no clock is a standing credential). A report grouped by an identifying field cannot be published at all, checked here and again at every render. Take one down with delete_reports_shares_id. POST /v1/reports/views/:id/shares (access: workspace:admin)

post_servicing-authorizations_id_accept

ChatGPT
Record the merchant's consent to debit Records that the customer agreed: the authorisation becomes live permission to debit this schedule. Body: consentToDebit, achAuthorizationAccepted and termsAccepted must ALL be true and sourceReference must name the funding source, or the database refuses it — an acceptance missing one of those is not an acceptance. The caller's IP and user agent are stamped from the request as the evidence of who agreed and from where. Legal from pending and not yet expired, so a second acceptance or a late one changes nothing and answers not found. Accepting is permission only: collecting the money is post_payments_id_collect, and recording money that arrived is post_payments_id_record. POST /v1/servicing-authorizations/:id/accept (access: workspace)

post_servicing-authorizations_id_cancel

ChatGPT
Withdraw an unanswered authorisation request Takes back a debit authorisation the customer never answered. Body: reason is required. Legal from pending only — once the customer has agreed, ending it is post_servicing_authorizations_id_revoke. A request that simply ran out of time is retired to expired on its own the next time the schedule's authorisations are read. POST /v1/servicing-authorizations/:id/cancel (access: workspace)

post_servicing-authorizations_id_revoke

ChatGPT
Revoke a live debit authorisation Ends a permission to debit that the customer HAD agreed to — they withdraw consent, or the desk stops debiting. Body: reason is required. Legal from accepted only; a request the customer never answered is withdrawn with post_servicing_authorizations_id_cancel instead, and the two are different words because they answer different questions about whether the customer ever agreed. Revoking does not pause collection on the schedule itself — that is post_payment_schedules_id_pause. POST /v1/servicing-authorizations/:id/revoke (access: workspace)

post_servicing_delinquencies

ChatGPT
Open a delinquency case Opens a collections case against a deal that is behind. Body: definitionKey and rowId name the record — both required — plus severity (early, moderate, severe, critical), daysDelinquent, missedPaymentsCount, missedAmount, totalOutstanding (decimal strings) and internalNotes. A deal already carrying a live case answers 409: two open rows on one deal is a double count, not a second problem. There is no detector behind this — a person states it, or post_servicing_delinquencies_sweep finds them from the ledger. POST /v1/servicing/delinquencies (access: workspace)

post_servicing_delinquencies_id_contact

ChatGPT
Log chasing a delinquent borrower Records an attempt to reach the borrower about a case that is behind: the attempt count goes up, the stamp says how it was made and how it went, and the note is appended. Body: method is REQUIRED — a contact attempt nobody can account for is refused — plus result and notes. Reach for post_servicing_delinquencies_id_resolve when the chasing is over. POST /v1/servicing/delinquencies/:id/contact (access: workspace)

post_servicing_delinquencies_id_resolve

ChatGPT
Resolve a delinquency case Ends a collections case. Body: status is a terminal word — cured when the deal came good, written_off, or settled — plus amount for what was recovered and notes. The resolution date is written in the same statement, so a resolved case with no date cannot exist. This closes the CHASING; how the deal itself ended in money terms is post_servicing_outcomes. POST /v1/servicing/delinquencies/:id/resolve (access: workspace)

post_servicing_delinquencies_sweep

ChatGPT
Find delinquent facilities now Runs the delinquency detection pass over this workspace now: reads the Payments board's own overdue view and keeps one delinquency row per facility in step with it, opening what is newly behind, moving the aging band (1-30 / 31-60 / 61-90 / 90+) and curing what came good. Answers the counts; GET /v1/servicing/delinquencies has the rows. The same pass the scheduled loop makes, so a person pressing it and the loop cannot disagree. POST /v1/servicing/delinquencies/sweep (access: workspace)

post_servicing_outcomes

ChatGPT
Record how a facility ended Closes a funded deal out in money terms: the payoff, the write-off, the settlement. Body: definitionKey and rowId name the record and outcomeType says how it ended (paid_in_full, early_payoff, default, write_off, settlement, refinanced, renewed, charged_off, bankruptcy) — all three required — plus outcomeDate, the money and rates as decimal STRINGS (totalFunded, totalPaybackExpected, totalCollected, totalOutstanding, totalWrittenOff, totalFeesCollected, totalPenalties, netProfitLoss, settlementAmount, settlementDiscountPct, collectionRate, the two rates between 0 and 1), daysToOutcome, totalDelinquentDays, reason and internalNotes. A second outcome on the same record answers 409 — a facility ends once. Stopping COLLECTION is post_payment_schedules_id_complete; this is the accounting of how it ended. POST /v1/servicing/outcomes (access: workspace)

post_signing-requests_id_cancel

ChatGPT
Cancel a signing request Withdraw a document's signing request, which stops every outstanding link working immediately. Reach for this when a paper was sent in error or has been superseded. It refuses on a request already executed or cancelled, and on a deal CONTRACT envelope - those close by voiding or cancelling the contract itself. POST /v1/signing-requests/:id/cancel (access: workspace)

post_signing-requests_id_remind

ChatGPT
Remind a request's signers Chase everyone on a signing request who has not yet signed or declined: each gets a NEW single-use link by email and their reminder count goes up. Reach for this when a merchant has gone quiet. The old links stop working, because this service stores only a hash of a token and cannot re-send one it does not hold. POST /v1/signing-requests/:id/remind (access: workspace)

post_sms-blasts_id_retarget-preview

ChatGPT
Count who a text blast would be resent to A QUESTION, not a send: how many of one text blast's recipients ended up in the outcomes you name, and a sample to recognise them by. Body: outcomes (any of queued, sent, delivered, failed, skipped), includeNoOutcome (addresses the dispatcher never spoke about at all, which is not the same as failed, so it is its own flag), sampleLimit. Each address's LATEST outcome decides, so a redelivery does not count twice. Nothing is created and nothing is queued — post_sms-blasts_id_retarget-send is what sends again. POST /v1/sms-blasts/:id/retarget-preview (access: workspace)

post_sms-blasts_id_retarget-send

ChatGPT
Send a blast again to whoever did not reply Sends a follow-up to everybody who did not ANSWER the first SMS blast — "did not answer" being the thread's own verdict: no live inbound reply in that address's thread since the copy that went out. It creates a NEW blast, gives it the non-responders as its audience, freezes it and dispatches it in one transaction, so a retarget that cannot finish leaves no half-built blast behind. The follow-up keeps the first blast's goal and each recipient's merge values, so its words may use the same fields and either/ors. Body: body (the follow-up's own words; absent copies the first blast's frozen ones), subject (email kinds only, refused on an SMS blast), medium ("sms" or "imessage"), limit (bounds one run) and pacing (when the copies may leave; a pacing this service cannot read is refused with a sentence, never quietly ignored). post_sms-blasts_id_retarget-preview asks the outcome question instead and creates nothing. POST /v1/sms-blasts/:id/retarget-send (access: workspace)

post_submissions

ChatGPT
Send one deal to one lender Sends ONE deal to ONE lender. rowId is the DEAL's id, or a record on this workspace's Submissions page: from a deal the server finds that deal's submission for this lender or makes it, and reads the merchant off the deal's Merchant, so merchantKey is needed only when the deal has none. The lender is lenderId (a record on your Lenders board, required when rowId is a deal) or lenderTenantId (the receiving workspace). The lender's record decides how it goes: linked to a Specter workspace (put_lender-workspaces_rowId) it is a share; otherwise the package goes through the submission mailer with the deal's files named in documentIds attached, under the workspace's standard template with note added; with neither it is refused 422 lender_unreachable. It is the same send as the deal's prepare-send and is refused for the same reasons: a mailed lender needs its PDF minimum among documentIds (4 unless its record says otherwise), every file landed and readable, an address that has not opted out, and submission email switched on for the workspace. 409 when this lender already has it. For several lenders in one call reach for post_submissions_bulk; to see where the sends already stand, get_submissions. Many at once: post_submissions_bulk POST /v1/submissions (access: workspace)

post_submissions_bulk

ChatGPT
Submit a deal to several lenders at once Submit one deal to up to 25 lenders in one call: each lender is routed and gated the way a single send is (a share to a linked Specter workspace, the package with the files in documentIds attached through the submission mailer to an address), each gets its own submission record, and each line says sent or refused, with the refusal's stable code beside its sentence. Use it for "send this deal to these lenders". One at a time: post_submissions POST /v1/submissions/bulk (access: workspace)

post_submissions_id_transition

ChatGPT
Update a lender submission share status Move one lender's crossing along - :id is the SHARE, not the submission, because one submission sent to four lenders has four independent states. The lender may move it to under review, approved or declined; the ISO may withdraw it. The move stamps its own instant (first response, decision, withdrawal), carries the reason and the decline code, and takes the figures the lender answered with in offered and counter. A lender's decline also raises a risk signal on the deal. POST /v1/submissions/:id/transition (access: workspace)

post_tasks

ChatGPT
Create a task Raises a task: title, an optional description, dueAt (RFC3339), priority, assigneeAccountId (the roster's account id), tags, and entityRef "<definitionKey>/<rowId>" to file it on a record. A task with a due TIME also lands on the assignee's calendar when their preferences mirror tasks. To book a meeting AND a task in one write, use post.calendar.events with createTask instead. Nothing is defaulted or de-duplicated: with no assigneeAccountId or dueAt the task is unassigned and undated, and the same task raised twice is two tasks — get_tasks with entityRef lists a record's tasks. entityKind alone names no record and is refused — a task about a record carries its entityRef. POST /v1/tasks (access: workspace)

post_tasks_id_comments

ChatGPT
Comment on a task Adds one comment to a task in the caller's name. A blank body is refused. POST /v1/tasks/:id/comments (access: workspace)

post_tasks_id_complete

ChatGPT
Mark a task done Closes one task in a single step — "mark it done", "that one is finished" — and answers the task as it now stands. The meeting it mirrors on the calendar is settled the way the assignee's calendar settings say. To close it with a cancellation reason, or to reopen it, change its status with PATCH /v1/tasks/:id instead. POST /v1/tasks/:id/complete (access: workspace)

post_undo_last

ChatGPT
Undo my last change Takes back the signed-in person's latest change in this workspace — whatever it was: a cell edit, a created or deleted record (it comes back with its trail, groups and follows), a bulk convert, a setting — or, with ?op=<id> from get.undo.history, that one. Call it ONLY when the person asks to undo or reverse something; never on your own initiative. Answers {undone, op, label, notUndone}: notUndone names what already left and cannot be called back (texts or emails sent, changes already synced, notifications already read) — say those out loud. 404 when there is nothing of theirs to undo; 409 undo_refused with a sentence when something it changed has been changed again since — the undo never overwrites a later edit, so tell the person which change to undo first. POST /v1/undo/last (access: workspace)

post_webhooks_id_test

ChatGPT
Queue a test webhook delivery Queue a webhook.test event for one subscription. It answers 202 and never makes an HTTP request — a separate worker delivers — so success here means queued, not received; read the outcome from the subscription's deliveries. Refused with 409 when the subscription is paused or does not subscribe to webhook.test, rather than queueing into silence. Admin only. POST /v1/webhooks/:id/test (access: workspace:admin)

post_workflows_id_inputs

ChatGPT
Set a shipped rule's values and turn it on TAKES EFFECT IMMEDIATELY: writes the values a pack rule asks for (its needs.inputs, such as a standard referral percentage) into the step each one names, then publishes the rule exactly as post_workflows_id_publish does — validated first, refused 422 with its problems rather than published broken. Body: values, an object of input key to value, carrying every input the rule declares and no other; a percent is a number from 0 to 100. Calling it on a live rule saves the new value and republishes. A rule whose step was edited in the builder so it no longer writes one cell is refused 409: change it there. POST /v1/workflows/:id/inputs (access: workspace:admin)

post_workflows_runs_id_cancel

ChatGPT
Stop a running workflow run Stops one run that is still going, at its next step. The steps it already took STAND — nothing it did is undone. This ends one execution, not the rule: a rule that should stop firing at all is archived through patch_workflows_id. POST /v1/workflows/runs/:id/cancel (access: workspace:admin)

post_workflows_validate

ChatGPT
Check a workflow definition Answers whether a workflow definition would run — valid plus a list of problems, each with a code and the step it is about. It CHANGES NOTHING and publishes nothing, so this is the safe way to ask "is this rule right". Body: definition as {trigger, steps}. It is the same code post_workflows_id_publish refuses on, and catches an unknown action or event, a duplicate or missing step id, a jump to nowhere, a backward jump, an expression or template that will not parse, a {{steps.x}} naming a step that runs later, and wait / review / stop / each nested inside an each. POST /v1/workflows/validate (access: workspace)

post_workspace_compliance_reports

ChatGPT
Generate a compliance report Admin only. A report is a name and the set of checks it runs; a report that asks nothing is refused. reviewWindowDays (90 by default) is how long a sign-off counts for and is read by the seat-review check. A second report under the same name is refused as a conflict. POST /v1/workspace/compliance/reports (access: workspace:admin)

post_workspace_compliance_reports_id_run

ChatGPT
Run a compliance report now Admin only. Evaluates every check the report names in one transaction and answers the run together with its findings. A run cannot be edited or re-run — asking again writes a new run — so its counts, its outcome and the checks it ran are fixed at this moment. The first run of a report carrying the seat-review check fails it for every seat, because nothing has been signed off yet; signing that run off is the review, and the next run passes. POST /v1/workspace/compliance/reports/:id/run (access: workspace:admin)

post_workspace_compliance_runs_runId_decisions

ChatGPT
Sign off, reject or reopen a run Admin only, and an APPEND: signing off a run that is already signed off adds a second line rather than overwriting the first, because two admins signing the same quarter is a fact worth keeping. decision is signed_off, rejected or reopened, and rejecting must say why. The signer's name is recorded as it stands now, so the trail still reads correctly after the person is renamed. The database refuses any edit to what is already there. POST /v1/workspace/compliance/runs/:runId/decisions (access: workspace:admin)

post_workspace_delegation_sync

ChatGPT
Synchronize delegated mailboxes and inboxes Walk the delegated directory and mint an inbox for every mailbox it finds, answering how many were found, minted, already there and now watched. The provider comes from the body or the ?provider= query. Reach for it after connecting a domain, or when people have joined it since. Admin and Enterprise only. POST /v1/workspace/delegation/sync (access: workspace:admin)

post_workspace_directory_invite

ChatGPT
Invite people from the directory Mints one invite link per address named, for addresses the connected directory actually holds, at the seat given or the workspace's configured default. Per address it answers what happened - invited, seated, invited_already or not_in_directory - so nothing is double-invited, and the links come back in links, the one and only place each token exists. Reach for it after reading the directory to let named people in. POST /v1/workspace/directory/invite (access: workspace:admin)

post_workspace_directory_sync

ChatGPT
Sync seats against the directory Applies the workspace's provisioning setting for one provider in a single sweep: invites everyone the directory holds who has no seat here (only when auto-invite is on - otherwise note says how many were found and skipped), and removes the seat of everyone the directory has suspended. Idempotent; running it twice invites nobody new. Reach for it after people join or leave the company. POST /v1/workspace/directory/sync (access: workspace:admin)

post_workspace_domains

ChatGPT
Add a workspace domain Claims a domain for the workspace and answers the DNS records to create: an ownership TXT that must be added before the domain verifies, plus SPF, DKIM where the connected mail provider has one, and DMARC. Nothing is resolved by this call - use the recheck action once the records are published. POST /v1/workspace/domains (access: workspace:admin)

post_workspace_domains_id_inbound

ChatGPT
Add an inbound address Creates an address at a verified domain and says where mail arriving there goes: a pack page or board (an unmatched message opens a new row) or a member (the thread becomes theirs). Refused with 409 if the domain is not verified yet, or if the address is already receiving somewhere. POST /v1/workspace/domains/:id/inbound (access: workspace:admin)

post_workspace_domains_id_recheck

ChatGPT
Recheck a domain's DNS Resolves the domain's records now and updates each record's result and the domain's status: finding the ownership record verifies it, losing it marks the domain failing. When the workspace serves its forms from an address under this domain, that address's CNAME is checked too and sets only the formHost's status (ok or failing), never the domain's. Rate limited to once every 30 seconds per domain; a second call inside that window answers 429 with Retry-After. POST /v1/workspace/domains/:id/recheck (access: workspace:admin)

post_workspace_drive-watches

ChatGPT
Watch a Drive folder Start watching a Google Drive folder, and say where its new files should land: {folderId, targetKind, ...}. targetKind is one of definition_row (with definitionKey + rowId), board_row (with boardRowId) or workspace (the document library, no record at all). An optional documentType stamps every document the folder lands. provider is google_drive (the default) or onedrive; a OneDrive folder (from GET /v1/workspace/onedrive-folders, with driveId for a SharePoint library) is read with the caller's own Microsoft connection before the watch exists, and the sweep reads it with that same connection. One watch per folder per workspace - a second is a 409. Watching starts immediately; the first sweep runs within a few minutes or on demand. POST /v1/workspace/drive-watches (access: workspace:admin)

post_workspace_drive-watches_watchId_sweep

ChatGPT
Sweep a Drive watch now List the watched folder right now and land every file this workspace has not already taken out of it, on the watch's target. Answers what it did: how many files the folder held, how many became documents, how many were skipped because they were already landed, and a note naming anything that could not be carried (a file over the size cap, a folder longer than one pass). The same work the scheduled sweep does, so pressing this cannot give a different answer. POST /v1/workspace/drive-watches/:watchId/sweep (access: workspace:admin)

post_workspace_heartbeat_actions_bulk

ChatGPT
Approve or reject pending automated follow-up drafts Accept or reject up to 50 pending heartbeat items at once. Send {"ids": [1,2], "accept": true} - accept is never assumed - with an optional reason for a rejection. Each id is settled on its own, because accepting sends real messages and no rollback could unsend them, so the answer carries a line per id saying whether it landed and why not. One at a time: post_workspace_heartbeat_actions_id_apply POST /v1/workspace/heartbeat/actions/bulk (access: workspace)

post_workspace_heartbeat_actions_id_apply

ChatGPT
Approve and send an automated follow-up draft Say yes to one planned action and let it happen — today that means sending the follow-up it drafted on its conversation. 409 when the action is no longer in the planned state, or is a kind with nothing to apply. If the thread's contact changed between drafting and approval the approval becomes a REFUSAL carrying that reason rather than a message to the wrong person. An admin decides any draft; a rep decides a draft on their own lead, and anybody else's reads as a 404. Many at once: post_workspace_heartbeat_actions_bulk POST /v1/workspace/heartbeat/actions/:id/apply (access: workspace)

post_workspace_heartbeat_actions_id_defer

ChatGPT
Snooze a pending automated follow-up suggestion Put one pending heartbeat suggestion off until a moment, so it leaves the queue and the record strip until then. The item stays planned - nobody has decided anything - it is simply not due. Send {"until": "2026-09-14"} or an RFC3339 timestamp; send nothing to snooze by the workspace's default. Refuses a moment in the past, one past the workspace's defer limit, and any item already settled. POST /v1/workspace/heartbeat/actions/:id/defer (access: workspace)

post_workspace_heartbeat_actions_id_refuse

ChatGPT
Reject a pending automated follow-up draft Say no to one planned action, with an optional reason kept beside it. The item is settled rather than deleted, so the run still reads as what was proposed and what was decided. Reach for defer instead to put an item off without deciding on it. An admin decides any draft; a rep decides a draft on their own lead. Many at once: post_workspace_heartbeat_actions_bulk POST /v1/workspace/heartbeat/actions/:id/refuse (access: workspace)

post_workspace_heartbeat_actions_id_revert

ChatGPT
Reopen a refused or failed automated follow-up Return a refused or failed heartbeat item to the queue while the workspace's take-back window is open. An APPLIED item cannot be reverted - the message has already gone out - and a window that has closed or is set to zero answers 410. Use it as the undo behind a refusal somebody pressed by mistake. POST /v1/workspace/heartbeat/actions/:id/revert (access: workspace)

post_workspace_invites

ChatGPT
Invite someone to the workspace Mint an invite link for a named role (owner, admin, team_lead or member); the app composes the sign-in link around the token, which appears in THIS answer and nowhere else. Admin only, and an owner-role invite is an owner's to mint. Accepting is POST /v1/invites/accept, which any signed-in account may call because the acceptor has no seat here yet. POST /v1/workspace/invites (access: workspace:admin)

post_workspace_members_accountId_pause

ChatGPT
Pause a member's seat Turn one member's access off WITHOUT removing them: the seat keeps its role, its teams, its assignments and everything the person ever wrote, and the door stops letting them in (their next call answers 423). Reach for this when somebody is on leave, between devices, or under review — anything where a removal would be too much and a demotion would not stop them. Admin only; an owner can only be paused by an owner, nobody may pause their own seat, and the last owner who is still live cannot be paused (409). Resuming is one call and restores nothing, because nothing was taken. POST /v1/workspace/members/:accountId/pause (access: workspace:admin)

post_workspace_members_accountId_resume

ChatGPT
Let a paused member back in Undo a pause: the seat is live again from the member's next call, with the role, teams and work it already had. Answers the member as it now stands. Resuming a seat that was never paused is not an error. Admin only, and an owner's seat only by an owner. POST /v1/workspace/members/:accountId/resume (access: workspace:admin)

post_workspace_processor_connections_mailbox_aliases

ChatGPT
Add an alias to the workspace's intake mailbox Adds an address to one of the workspace's processor accounts: a plus-address of it, an address Gmail already delivers to it, or with create: true a new alias made at Google through the workspace's domain-wide delegation, on a domain the workspace has verified. Google's refusal comes back in Google's words with what to do. POST /v1/workspace/processor/connections/:mailbox/aliases (access: workspace:admin)

post_workspace_processor_criteria_test-run

ChatGPT
Test the processor criteria on recent mail Judges the saved criteria — or a draft sent in the body, validated like a save — against this workspace's last 50 messages at processor@, and answers per message what became of it then, what would become of it now, the intake rule that would decide it and the completeness rules' results. Nothing is written and nobody is emailed. Admin only. POST /v1/workspace/processor/criteria/test-run (access: workspace:admin)

post_workspace_processor_file-types_key_examples

ChatGPT
Teach a file type with an example Uploads one document ({name, contentBase64}) to processor-v2 as an example of a file type; 202 with the example, status reading until processor-v2 has read it in the background — list the examples again to see it settle. The same file sent twice is one example, read afresh. :key is the app's document word or a workspace's own custom_<slug> type; 400 for a word processor-v2 has no type for, 503 when this deployment has no processor-v2. Admin only. POST /v1/workspace/processor/file-types/:key/examples (access: workspace:admin)

post_workspace_processor_whitelist

ChatGPT
Let a seatless sender email the processor Adds one domain (acme.com) or address (dana@acme.com) this workspace lets email processor@ without a seat, with a required reason. Their deals are made as the seat of the admin adding the entry. Values are lowercased; a sender the workspace already lets in is a 409. Admin only. POST /v1/workspace/processor/whitelist (access: workspace:admin)

post_workspace_rep_thread_messages

ChatGPT
Create workspace rep thread messages Send a message to the workspace's representative, opening the thread if this is the first one. 404 when no rep is assigned. ANY SEAT, deliberately: the person with the question is the one who should be able to ask it, and routing it through an admin loses them on the way. POST /v1/workspace/rep/thread/messages (access: workspace)

post_workspace_role-presets

ChatGPT
Create a role from a preset Copies one preset (fromPresetId) into the firm's own under a new name, optionally with a description and, for a staff preset, another seat (member, team_lead, admin). The copy keeps the source's board levels, books level and, for an affiliate, its reach. Answers the new preset. A workspace admin's. POST /v1/workspace/role-presets (access: workspace:admin)

post_workspace_sidebar-folders

ChatGPT
Make a sidebar folder Makes a folder named label at the end of the rail, or, when the workspace already has a folder by that name (casing and padding aside), answers that one with 200 instead of 201. Either way the optional targets (route words, or "board:<board id>") are filed at the folder's foot, moving out of any other folder: "put this board in the Nonprofit folder, made if missing" in one call. A board the caller cannot open is refused as not found. Any seat. POST /v1/workspace/sidebar-folders (access: workspace)

post_workspace_sla-policies

ChatGPT
Promise a stage an SLA Admin only. Names the pack page and the stage band it measures, the target and warning in minutes (the warning must be the smaller), and optionally a business-hours calendar: an IANA time zone, the working day as minutes from midnight, and the working days as 0 (Sunday) through 6 (Saturday). A second policy on the same page and stage is refused as a conflict, and a time zone this service cannot load is refused with the name in the message. POST /v1/workspace/sla-policies (access: workspace:admin)

post_workspace_support_tickets

ChatGPT
Open a support ticket Open a support ticket with us: a subject and a body, both required, and nothing else — status and priority are ours to set, because a raiser who can choose 'urgent' chooses it every time. ANY SEAT on purpose: the member whose board will not load is the person who saw it. A workspace with an assigned representative has the ticket start with that person's name on it. POST /v1/workspace/support/tickets (access: workspace)

post_workspace_support_tickets_id_comments

ChatGPT
Comment on a support ticket Add the caller's reply to one of their own tickets. There is no internal flag on this route at all rather than one that is ignored — a customer's comment can never be internal — and offering the key would suggest otherwise. Any seat. POST /v1/workspace/support/tickets/:id/comments (access: workspace)

post_workspace_teams

ChatGPT
Create a team Create a team, optionally naming its lead by account id. Admin only. 409 when a team of that name is already here. POST /v1/workspace/teams (access: workspace:admin)

post_workspace_underwriting-criteria_seed

ChatGPT
Seed the workspace's underwriting criteria Install the standard scorecard, adding only the criteria whose key is not already there and answering how many were created. Pressing it twice creates nothing and files nothing on the activity feed. Admin only. Reach for it on a workspace with an empty scorecard rather than writing the usual questions out by hand. POST /v1/workspace/underwriting-criteria/seed (access: workspace:admin)

post_workspace_underwriting_rule-sets

ChatGPT
Create an underwriting rule set Create a DRAFT rule set: a name, and optionally the deal stage and product it narrows to (both empty mean every deal). It judges nothing until it is published. Add rules with the rules route, then publish. POST /v1/workspace/underwriting/rule-sets (access: workspace:admin)

post_workspace_underwriting_rule-sets_seed

ChatGPT
Seed the default underwriting rule set Write the default credit box - twelve rules over bank statements, tax returns, credit reports and the deal's own cells - and publish it. Idempotent: a second call answers the set already there with created=false, and never overwrites an edit. Reach for it when a workspace has no rule set yet. POST /v1/workspace/underwriting/rule-sets/seed (access: workspace:admin)

post_workspace_underwriting_rule-sets_setId_publish

ChatGPT
Publish an underwriting rule set Freeze the set's active draft rules as the next version and point the set at it, so evaluations from now on run those words. Refused when the set is archived or has no active rule, and when another live set already claims the same stage and product. POST /v1/workspace/underwriting/rule-sets/:setId/publish (access: workspace:admin)

post_workspace_underwriting_rule-sets_setId_rules

ChatGPT
Add an underwriting rule Add one rule to a set's draft: a condition in the formula grammar over the namespace (stmt., tax., credit., extract., deal.*), an outcome for when it does not hold (pass, fail or review), a reason template and a weight. The condition is parsed here, so a malformed one is refused rather than erroring on every deal later. Publish to make it run. POST /v1/workspace/underwriting/rule-sets/:setId/rules (access: workspace:admin)

post_workspace_vendors

ChatGPT
Engage a vendor Admin only. name and kind are required, and kind is processor, attorney, credit_bureau, notary or other. Contract dates are YYYY-MM-DD, or null when they are not on file, and an end date before its start is refused. A second vendor under the same name, case-insensitively, is refused as a conflict. POST /v1/workspace/vendors (access: workspace:admin)

post_workspace_vendors_id_documents

ChatGPT
Pin a document to a vendor Admin only. Names a file already in this workspace's library - the engagement letter, the W-9 - and attaches it to the vendor. The file is not uploaded or copied here: a pin is a reference, so one document keeps one storage key. Pinning what is already pinned is the same pin. POST /v1/workspace/vendors/:id/documents (access: workspace:admin)

preview_blast_audience

ChatGPT
For non-merchant Specter broker/operator users only. Counts how many LEADS match a set of filters — status, tags, owner, group — and returns a small sample, without sending anything. Use it to answer "how many leads would this reach" before planning outreach, and state the count to the user: these filters routinely resolve to a different set than the person pictured, and exclude_do_not_contact and exclude_recently_sent move the number most. You must pass at least one POSITIVE selector — lead_ids, statuses, group_ids, owner_ids, tags or search. A call carrying only exclusions or has_email/has_phone comes back with no_selection set and a count of 0, which means you selected nobody, NOT that the pipeline is empty; say so rather than reporting there are no leads. The ids it returns are LEAD ids from the leads table; do not pass them to a tool that expects CRM record ids, which are a different set entirely. It is read-only and sends nothing.

preview_blast_retarget

ChatGPT
For non-merchant Specter broker/operator users only. Counts who a re-send of an earlier SMS blast would reach, and sends NOTHING. Always call this before send_blast_retarget and tell the user eligible_count — that number is the whole point of the preview, and there is no recipient ceiling on the send. Pass the SAME target to both calls: preview and send each recompute the audience independently, so a different target sends to a different set than the one you showed. Recipients already retargeted from this blast are excluded, which is why a second preview of the same target can legitimately come back with zero. It sends nothing, but it is not inert: it refreshes each recipient's reply and opt-out status from inbound SMS as a side effect.

propose_meeting

ChatGPT
For non-merchant Specter broker/operator users only. Drafts a calendar meeting tied to a lead, merchant (counterparty) or deal (facility) as a PENDING Heartbeat action. NOTHING IS SENT by this call: the invite goes out only when a person accepts the action (in the Heartbeat review queue, or accept_heartbeat_action + execute_heartbeat_action after the user explicitly says so). Before calling, confirm with the user the time (an RFC3339 start_time with offset), the duration, and WHO is invited: the record's contact is invited unless include_contact is false; set include_assigned_users to also invite every user assigned to the record; name anyone else under attendees. The organizer is the record's owner (deal opener, lead owner, merchant assignee) and it lands on their connected calendar. Optionally tie the meeting to a task: task.task_id links an existing task, task.create makes one on accept (due at the meeting by default, assigned to the organizer). Answers the pending action with its id; tell the user it is waiting for their accept.

publish_workflow

ChatGPT
Publishes a Workflow V3 draft for the authenticated tenant. The backend re-validates the STORED draft graph at publish time and refuses an invalid one, so for an already-saved draft the assistant may publish by workflow_id even when the graph is not in this conversation (use get_workflow first only when the user wants the contents summarized or changed). Use after the user explicitly confirms publishing. This activates the stored automation: later matching triggers can run its configured record changes and outbound communications under the tenant's permissions. It does not register or expose additional model-callable tools. The input requires confirm_publish=true.

put_accounting_access_accountId

ChatGPT
Grant a seat the books Hands one member the brokerage's books, or changes what they hold. Body: level (read opens the books; manage also books deals, lays stakes, reconciles bank lines and records costs), allAccounts (true reaches every linked company bank account and hand-imported lines), and bankAccounts — when allAccounts is false, the linked accounts the grant reaches, by the id each account carries in get_accounting_plaid-connections. The whole grant is replaced by the body. A grant never decides a payout or links a bank; those stay the admin seat's. The admin seat. PUT /v1/accounting/access/:accountId (access: workspace:admin)

put_accounting_bank-transactions_id_classification

ChatGPT
Categorize and tag a transaction Assign a saved category identifier and tags to a selected-account bank transaction. Read the transaction first, pass detail.classification_revision as expectedRevision (zero if absent), and preserve tags when changing only the category. An empty categoryId clears the category; JSON null is also accepted. An optional note (at most 2000 characters) sets the line's own note; leave it out to keep the existing note, send "" to clear it. A stale revision returns a conflict; reload before retrying. This classification does not move money or change allocations and supports undo. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. A grant scoped to some linked bank accounts reaches only those accounts' lines. PUT /v1/accounting/bank-transactions/:id/classification (access: workspace:books)

put_accounting_categories_identifier

ChatGPT
Save a transaction category Create or update a workspace category. Use a stable lowercase identifier; renaming the display name preserves that identity. Guidance and keywords explain how Neo should classify expenses and income. active=false archives a category without erasing prior classifications. Changes support undo. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. PUT /v1/accounting/categories/:identifier (access: workspace:books)

put_accounting_category-order

ChatGPT
Reorder transaction categories Set the order categories are listed in everywhere (pickers, the category ring, the manager). Send identifiers in the order they should appear; any category left out keeps its relative place after them. An unknown or repeated identifier is refused. Returns the categories in their new order. Takes the brokerage's books: an owner or admin, or a seat holding an accounting grant (get_accounting_me_access); anyone else is refused with 403. PUT /v1/accounting/category-order (access: workspace:books)

put_accounting_company-accounts_plaid_plaidAccountId

ChatGPT
Set a linked account's purpose Gives a linked COMPANY bank account (Accounting › Settings › Linked banks) its purpose and practice areas, or changes them. Body: purpose, packKeys. A merchant's bank on a deal is never accepted. Takes the admin seat (owner or admin); anyone else is refused with 403. PUT /v1/accounting/company-accounts/plaid/:plaidAccountId (access: workspace:admin)

put_accounting_invoice-settings

ChatGPT
Save the workspace's invoicing setup Saves any of businessName, logoUrl (https), accentColor (#rrggbb), numberPrefix, nextNumber (above every number already printed with that prefix), termsDays (0–365), taxPpm, paymentInstructions; the automatic reminders: remindersOn (on by default), reminderDays (days from the due day a reminder goes, negative before — default [-3, 0, 3, 7, 14]; at most 8, from -30 to 90), callAfterDays (the day past due the invoice's owner gets a task to call, default 15; 0 for none); complete: true marks the guided setup finished, which also turns on the workspace's invoice follow-up rules. New drafts start from these. Answers the setup as get.accounting.invoice-settings does. Takes the books to KEEP: an owner or admin, or a manage accounting grant; anyone else is refused with 403. PUT /v1/accounting/invoice-settings (access: workspace:books)

put_accounting_invoices_id

ChatGPT
Save a draft invoice Changes a draft's terms (billToName, billToEmail, memo, paymentInstructions, currency, taxPpm, dueOn) and, when lines is sent, replaces every line. An issued invoice is frozen and refused with 409: void it and issue a new one. Takes the books to KEEP; anyone else is refused with 403. PUT /v1/accounting/invoices/:id (access: workspace:books)

put_accounting_plaid-connections_itemId_accounts

ChatGPT
Select company accounts for the ledger Replaces the complete set of selected Plaid account IDs and imports their stored posted history. An empty list stops importing all accounts on this bank; existing ledger entries remain. PUT /v1/accounting/plaid-connections/:itemId/accounts (access: workspace:admin)

put_accounting_settings

ChatGPT
Save accounting settings Replace this workspace's accounting settings. Read them first and send the whole object back with your changes; a field left out keeps its stored value. Currency is three capital letters; the fiscal year starts in month 1-12; wait and due-soon days are 0-365; deal SLA and clawback window days are 1-365. terms maps vocabulary to key to {label, description, fix}: only known vocabularies and keys are accepted, a label is 1-60 characters, and any key not named keeps its default wording. New deals that omit currency, SLA or clawback window take these defaults. PUT /v1/accounting/settings (access: workspace)

put_boards_id_actions_order

ChatGPT
Reorder a board's action bar REPLACES the order the buttons are drawn in on one custom board. order must list every action id on the board. Manage on the board. PUT /v1/boards/:id/actions/order (access: board:manage)

put_boards_id_column-policies_column

ChatGPT
Lock a column on a board Names who may SEE one column's values on a custom board and who may CHANGE them, by audience. Takes the workspace ADMIN seat as well as manage on the board, and the Enterprise plan — the seat is checked first, so a member of a broker workspace hears 403 rather than an invitation to buy a control they still could not use. PUT /v1/boards/:id/column-policies/:column (access: workspace:admin)

put_boards_id_columns_columnId_calculation

ChatGPT
Choose the figure the calculations strip shows under a board column Sets the calculation the summary strip shows under one column of a custom board — sum, average, median, min, max, earliest, latest, count filled, count empty, percent filled, percent empty or count unique — for everyone who opens the board; an empty kind clears it. Any seat that can read the board may choose, since this is a view choice and not the column's shape. 400 for a word the strip does not have, 404 for a column the board does not carry. PUT /v1/boards/:id/columns/:columnId/calculation (access: board:read)

put_boards_id_columns_columnId_formula

ChatGPT
Set a board column's formula expression Replaces the expression on a formula column of a custom board and recomputes its stored values. Needs manage access on the board; 404 for a column the board does not carry, 400 for an expression the evaluator rejects. Reach for it after previewing with POST /v1/formulas/preview, when the person has decided what the column should compute. PUT /v1/boards/:id/columns/:columnId/formula (access: board:manage)

put_boards_id_columns_columnId_options_order

ChatGPT
Reorder a board column's choices REPLACES the order one select column's choices are drawn in. order must list every option id on that column. Manage on the board. PUT /v1/boards/:id/columns/:columnId/options/order (access: board:manage)

put_boards_id_columns_order

ChatGPT
Reorder a board's columns REPLACES the left-to-right order of a custom board's columns. order must list EVERY column id on the board — a short list is refused, never partly applied. Manage on the board. PUT /v1/boards/:id/columns/order (access: board:manage)

put_boards_id_grants_accountId

ChatGPT
Share a board with a colleague Gives one account read, write or manage on a custom board, or changes the level it already holds. Naming the board's own maker answers 409 — they already have everything on it. Manage on the board. PUT /v1/boards/:id/grants/:accountId (access: board:manage)

put_boards_id_views_order

ChatGPT
Reorder a board's saved views REPLACES the order the view tabs are drawn in for ONE scope — user (the default, the caller's own) or workspace (manage on the board). order must list every view id of that scope; the two scopes are separate lists and reorder separately. PUT /v1/boards/:id/views/order (access: board:read)

put_boards_id_workspace-access

ChatGPT
Set a board's workspace access Sets the level every seat with no grant of its own gets on a custom board. Body {"level": "manage"|"write"|"read"|"assigned"}; assigned is the default rung (a member sees the rows assigned to them). A grant still wins for its holder. Recorded on every plan, it governs access once the workspace is on Enterprise. Manage on the board. PUT /v1/boards/:id/workspace-access (access: board:manage)

put_comms_conversations_id_labels_labelId

ChatGPT
Label a conversation Files one thread the caller can read under a workspace inbox label. Labelling it twice is a no-op; a thread hidden from the caller is a not-found. PUT /v1/comms/conversations/:id/labels/:labelId (access: workspace)

put_comms_inboxes_medium_identifier_members_accountId

ChatGPT
Put somebody on an inbox Adds one account to the people on a number or a mailbox, which lets them read its private threads and send from it. Idempotent — 201 when the row is new, 204 when they were already on it — and 409 when the account holds no seat in this workspace. An admin for any inbox; the owner for their own connected mailbox or Zoom line. PUT /v1/comms/inboxes/:medium/:identifier/members/:accountId (access: workspace)

put_comms_inboxes_medium_identifier_visibility

ChatGPT
Set one inbox's privacy Sets whether new threads on one inbox are visible to the workspace or private to their owner and the inbox's members, and whether a colleague's own mailbox or Zoom line is open for everyone to send from. Answers the whole visibility bag. An admin for any inbox; the owner for their own connected mailbox or Zoom line. PUT /v1/comms/inboxes/:medium/:identifier/visibility (access: workspace)

put_comms_settings_visibility

ChatGPT
Set who may see messages by default Replaces the whole bag of visibility defaults, keyed by medium and by inbox. A key the body leaves out is DROPPED, which is exactly how a privacy pane turns one back off, so send the full bag. A key naming an inbox the workspace does not have is refused, quoting it. Admin only. PUT /v1/comms/settings/visibility (access: workspace:admin)

put_comms_signatures

ChatGPT
Set an address's email signature Saves the sign-off appended to every email leaving one address. The address is in the BODY, not the path. An address the workspace does not send from is refused with a 400 naming it, and one that is ours but not this seat's to sign with a 403. PUT /v1/comms/signatures (access: workspace)

put_connections_id_mail-folders

ChatGPT
Choose folders to scan Read or choose native Yahoo/AOL folders for your own connected workspace mailbox. Requires the current mailbox owner and active consent. Folder names and derived data are prohibited from advertising use. PUT /v1/connections/:id/mail-folders (access: workspace)

put_definitions_key_actions_order

ChatGPT
Reorder a page's row actions Sets the left-to-right order of a page's action-bar buttons. order must list EVERY action id on the page, and a partial list is refused with 'order must list every action id' rather than half-applied. Answers 204. Workspace admin. PUT /v1/definitions/:key/actions/order (access: def:admin)

put_definitions_key_child-pages

ChatGPT
Declare the child pages a page accepts Replace the whole ordered list of child pages this page will hold subrecords of. Admin only, because it is the shape of a shared surface. A page that stops being accepted takes its existing LINKS with it and never the child records; renaming a section re-titles the links already drawn in it. PUT /v1/definitions/:key/child-pages (access: def:admin)

put_definitions_key_column-policies_column

ChatGPT
Lock or unlock a column Locks one column, or replaces the lock on it — who may see it and who may write it, stated whole. A PUT rather than a patch because a policy is only meaningful entire: 'I cleared the editing restriction' and 'I did not mention it' must not look the same. A body that restricts NOTHING deletes the policy and answers the unrestricted one, so unrestricted has exactly one spelling. Workspace admin, and on the enterprise tier — the reads stay open on every plan so a policy already set keeps applying. PUT /v1/definitions/:key/column-policies/:column (access: def:admin)

put_definitions_key_columns_columnId_calculation

ChatGPT
Choose the figure the calculations strip shows under a page column Sets the calculation the summary strip shows under one column of a record page (leads, deals, merchants and the rest) — sum, average, median, min, max, earliest, latest, count filled, count empty, percent filled, percent empty or count unique — for everyone who opens the page; an empty kind clears it. Any member may choose, since this is a view choice and not the page's shape, which stays an admin's through PATCH. 400 for a word the strip does not have, 404 for a column the page does not carry. PUT /v1/definitions/:key/columns/:columnId/calculation (access: def:read)

put_definitions_key_columns_columnId_options_order

ChatGPT
Reorder a page column's choices REPLACES the order one select column's choices are drawn in. order must list every option id on that column. Workspace admin. PUT /v1/definitions/:key/columns/:columnId/options/order (access: def:admin)

put_definitions_key_columns_order

ChatGPT
Reorder a page's columns Sets the left-to-right order of a page's columns. order must list EVERY column id on the page — a partial list is refused with 'order must list every column id' rather than half-applied — and each column's position becomes its index in that list. Answers 204. Workspace admin. PUT /v1/definitions/:key/columns/order (access: def:admin)

put_definitions_key_grants_accountId

ChatGPT
Hold a member back on a pack page Sets one member's level on a pack page: none keeps its rows from them everywhere, read lets them see but not change, write is the default. Naming an owner or admin answers 409. Admin seat, Enterprise. PUT /v1/definitions/:key/grants/:accountId (access: def:admin)

put_definitions_key_rows_rowId_assignee_role

ChatGPT
Put one person in a role on a record Puts one member in one role on one DEAL by writing that role's person cell — opener or closer — and answers the row; an empty assignee clears the role. This is the single-cell route. put_definitions_key_rows_rowId_assignments is the one that replaces a role's whole SET, works on every pack page, and spells the deal's opener agent; both write the same column, so they cannot disagree. Deals only. Many at once: post_definitions_key_rows_assignees PUT /v1/definitions/:key/rows/:rowId/assignee/:role (access: def:read)

put_definitions_key_rows_rowId_assignments

ChatGPT
Set who holds a role on a record Replaces the WHOLE set of people in one role on one record and answers the diff — who was added and who was removed — so the caller need not work that out. accountIds order is meaningful: the first is the role's primary and lands in the record's person cell; an empty list clears the role. Roles are agent, closer, originating_agent and processing_agent on deals, owner on leads, submissions and merchants. Who a caller may put on or take off a record is checked for every id on both sides of the change. PUT /v1/definitions/:key/rows/:rowId/assignments (access: def:read)

put_definitions_key_rows_rowId_books_transactions_transactionId_category

ChatGPT
Categorize a client's bank line Files one synced line under an account of the record's chart (accountId; empty hands it back to the rules). remember: true also teaches a payee rule from the line's merchant (or description), so the next line from that payee files itself; a person's rule outranks every template rule and a second override of the same payee moves it. The line must be on one of this record's banks. Takes a Businesses or Clients record of an accounting firm's pack (a page whose template keeps books); any other page is a not-found. Reading needs the record; changing anything needs the seat that may edit the record's cells - a seat held to read the page is refused with 403. PUT /v1/definitions/:key/rows/:rowId/books/transactions/:transactionId/category (access: def:read)

put_definitions_key_rows_rowId_documents_documentType_standing

ChatGPT
Set a document's standing on a deal Sets ONE record's own standing on ONE document (0208): severity — required, warning (the app's "Needed") or nice_to_have ("Helpful"), or "" to hand the decision back to the deal type's rule — and waived with an optional waivedReason, which flags that paper not necessary on THIS deal. The document type is the path segment. Use it when a deal type's required documents do not fit a particular merchant (a sole trader with no articles of incorporation, a licence already on file from the last funding) — it changes nothing for any other deal, unlike editing the workspace's completeness rules. A waived item stays on the checklist and stops counting towards complete. PUT replaces, so saying it twice is saying it once and there is no 409. Answers the whole refreshed checklist, the same shape get_definitions_key_rows_rowId_documents_checklist returns. PUT /v1/definitions/:key/rows/:rowId/documents/:documentType/standing (access: def:read)

put_definitions_key_rows_rowId_email-rules

ChatGPT
Save lender email processing rules Replaces this lender record's workspace override object. Empty object restores inherited defaults. Explicit false and empty strings override defaults. Validates Go RE2 link patterns and preserves imported tenant approval instructions. Requires permission to edit the lender record. PUT /v1/definitions/:key/rows/:rowId/email-rules (access: def:read)

put_definitions_key_rows_rowId_feed_reactions

ChatGPT
React to a feed item Set the caller's reaction on one item of a record's feed, naming the item by itemKind and itemId from the feed read. A member holds at most one reaction per item, so sending a different emoji moves theirs rather than adding a second. Answers the item's whole reaction line. PUT /v1/definitions/:key/rows/:rowId/feed/reactions (access: def:read)

put_definitions_key_rows_rowId_files_fileId_folder

ChatGPT
File a document into a folder Move one of a record's files into a folder, or back to the record's Documents root by sending folderId as null. The folder must be on the same record. PUT /v1/definitions/:key/rows/:rowId/files/:fileId/folder (access: def:read)

put_definitions_key_rows_rowId_follow

ChatGPT
Follow a record Subscribe the caller to a record's feed, so they are notified when something new lands on it even though they do not hold the record. Following twice is the same subscription. PUT /v1/definitions/:key/rows/:rowId/follow (access: def:read)

put_definitions_key_rows_rowId_lender-form-fills_documentId_values

ChatGPT
Correct the values of a form issued on a record Keeps a seat's corrections to a form issued on this record: values, {fieldId: value} — text on one line, an amount as dollars, a date as YYYY-MM-DD, a checkbox true or false (ticking one box of a pick-one group clears the others). Any field but a signature, an initial or a signing date. The values frozen at issue and the signer's own answers are kept as they were; a correction is laid over them, named on the record's timeline, and drawn on the next printed copy. Refused on a form already signed electronically. Requires write access to the record. PUT /v1/definitions/:key/rows/:rowId/lender-form-fills/:documentId/values (access: def:read)

put_definitions_key_rows_rowId_lender-forms_formId

ChatGPT
Save a lender form's corrections Save the editor's fields (and optionally a new title) as the form's next version. Saving over a version that is no longer the newest is a 409. Each field: id, page, x/y/width/height (0..1 of the page from its top-left), kind (text, date, checkbox, signature, initials), label, word, option (the answer a checkbox stands for), signer (borrower, sponsor, guarantor or broker, for a signature, an initial or a signing date). Requires write access to the row. The fields a person confirmed also count as this workspace's one vote, per blank, for the catalog concept its word means, in the shared form library (never the word itself). PUT /v1/definitions/:key/rows/:rowId/lender-forms/:formId (access: def:read)

put_definitions_key_rows_rowId_lender-options

ChatGPT
Choose a submission's primary lender offer Selects a zero-based index from the displayed replyId. Refuses stale reply versions and closed submissions. Updates only submission terms and records the broker's choice; linked merchant offers remain unchanged and receive a review flag. Requires record write access. PUT /v1/definitions/:key/rows/:rowId/lender-options (access: def:read)

put_definitions_key_rows_rowId_plaid_monitoring

ChatGPT
Set merchant monthly monitoring Set enabled true or false for this linked deal, or null to inherit the admin default. Enrollment is separate from bank consent and does not disconnect banks. PUT /v1/definitions/:key/rows/:rowId/plaid/monitoring (access: def:read)

put_definitions_key_rows_rowId_submission-notes

ChatGPT
Save deal submission notes Persists reviewed lender notes on this deal with an expected version. Requires submission write access; stale versions return a conflict instead of overwriting another edit. PUT /v1/definitions/:key/rows/:rowId/submission-notes (access: def:read)

put_definitions_key_rows_rowId_vendor-slots_kind

ChatGPT
Name the vendor for one slot Seats a vendor in this record's slot for one kind. PUT because a slot has ONE occupant: filling a filled slot replaces it rather than adding a second. The vendor must be OF that kind and must not be retired - an attorney cannot fill the notary slot, and a retired vendor takes no new work. PUT /v1/definitions/:key/rows/:rowId/vendor-slots/:kind (access: def:read)

put_definitions_key_statuses_order

ChatGPT
Reorder a page's statuses Sets the order a page's statuses are drawn in — the order of its board columns and status pickers. order must list EVERY status id on the page; a partial list is refused with 'order must list every status id' rather than half-applied. Answers 204. Workspace admin. PUT /v1/definitions/:key/statuses/order (access: def:admin)

put_definitions_key_transition-forms

ChatGPT
Set required fields for a record status change or conversion REPLACES the field list for one trigger — a status value, or convert for the lead-to-deal handoff. Each field names a real column on this page and may be marked required, which refuses the move until it is filled. An empty list removes the form. PUT /v1/definitions/:key/transition-forms (access: def:admin)

put_definitions_key_underwriting_id_criteria

ChatGPT
Save manual underwriting scorecard answers Records a person's scorecard answers into the CURRENT version and answers that whole version, refreshed. Each entry names a criterionId and may carry status, numericValue and note; each field is independently optional, so a save carrying only a figure keeps the status beside it. An answer given here is stamped manual with who gave it, which is what makes a later evaluate leave it alone. A version is a PASS of the machine, not a keystroke, so this does not mint one. Deals page only. PUT /v1/definitions/:key/underwriting/:id/criteria (access: def:read)

put_definitions_key_workspace-access

ChatGPT
Set a pack page's workspace access Sets the level every member with no grant of its own gets on a pack page. Body {"level": "write"|"read"}; read makes the page read only for them, enforced on its rows by the database. A per-member grant still wins, and owners and admins always write. Admin seat, Enterprise. PUT /v1/definitions/:key/workspace-access (access: def:admin)

put_document-templates_forms_formId

ChatGPT
Save a document template form The editor's corrections to a workspace PDF form as its next version: fields (each a place on a page, 0..1 from the top-left, with the record word, the sender or the signing party that fills it — boxes the PDF never had, a choice group of checkboxes picking one or many with an option value each, an amount line printing the choice) and title to rename it. A save over a version that is no longer the newest is a 409. Admin only. PUT /v1/document-templates/forms/:formId (access: workspace:admin)

put_files_id_portal

ChatGPT
Show a file in the merchant portal Let the deal's merchant see one file of the deal in their portal, answering the file with sharedToPortal true. The portal already shows the merchant's own uploads and the files that answered an ask; everything else on the deal - the desk's own papers, a lender's mail - stays hidden until shared here. Only a file filed on a deal can be shown (400 otherwise). Sharing what is already shared is the same answer. PUT /v1/files/:id/portal (access: workspace)

put_files_id_star

ChatGPT
Star a file Star one file for the caller only — nobody else sees another person's stars. Answers {starred: true}; starring twice is the same answer. The files listing carries each file's starred for whoever reads it. PUT /v1/files/:id/star (access: workspace)

put_integrations_managed-zoom_draft

ChatGPT
Save managed phone setup Save preparation for a Specter-managed Zoom line, as workspace owner/admin. Read the current draft first and provide its revision; stale revisions return 409. Use facts supplied by the user, never invent company details or consent. Do not include EINs, credentials or payment details. This saves a draft only: no Zoom filing, purchase, Stripe charge or carrier approval occurs. PUT /v1/integrations/managed-zoom/draft (access: workspace:admin)

put_integrations_ringcentral_ai

ChatGPT
Set RingCentral AI processing Set enabled for workspace-AI transcription fallback and summaries on the caller's RingCentral recordings. Enabling permits processing through the workspace AI configuration and requeues eligible enrichment. Provider transcripts can still be imported while disabled. Answers 204. PUT /v1/integrations/ringcentral/ai (access: workspace)

put_lender-workspaces_rowId

ChatGPT
Link a lender record to its Specter workspace Links one entry on our Lenders board to the Specter workspace that lender actually IS, so deals can be submitted to them in-network instead of as e-mailed paper. rowId is the Lenders record; body: lenderTenantId, the lender's workspace. A PUT because the record is the link's identity — the same call makes the link and corrects it, and repeating it changes nothing. delete_lender-workspaces_rowId takes the link off and leaves the record standing. PUT /v1/lender-workspaces/:rowId (access: workspace)

put_mailboxes_id_intake

ChatGPT
Aim an inbox at the page new senders open on Sets what an email from a sender no record claims OPENS when it arrives at one connected inbox: routesTo: {kind: "definition", ref: <page key>} opens a new row on that page (pack page or made board) carrying the sender's name, address and signature phone, linked to the thread, with the attachments filed on it and the page's intake defaults, announced as a created record with source: email_intake; kind: "member" makes the thread that person's; routesTo: null stops the inbox opening anything. A reply, or a sender already on a record, still files to that record first. Answers the mailbox with intake. 400 for a page, board or person this workspace does not have. Admin only. PUT /v1/mailboxes/:id/intake (access: workspace:admin)

put_me_script-template

ChatGPT
Replace my call-script template Save the signed-in person's whole call script and tone. Send everything the read answered with the edits merged in — this replaces rather than patches, so a body carrying only the tone would blank the script. An empty spine, a tone picker value outside the vocabulary, or a slider outside 0-100 is refused with the field named. PUT /v1/me/script-template (access: workspace)

put_preferences_key

ChatGPT
Replace preferences key Saves one of the signed-in person's own app preferences. key must be namespaced, like setting:theme or table:leads; body: {"value": <json>}, stored opaquely and handed back unread, where a JSON null clears the preference and anything over 64KB is refused. It writes the caller's own row and nobody else's; the workspace-wide default under the same key is put_workspace_preferences_key, which is an admin's. PUT /v1/preferences/:key (access: workspace)

put_table-preferences_tableKey

ChatGPT
Replace table-preferences tableKey Saves how the signed-in person has arranged one grid. tableKey must be namespaced, like table:deals:overview; body: columnOrder and hiddenColumns as column ids (at most 200 each, and empty legitimately means unset — no order is the table's own order), sortColumn with sortDirection, and groupColumn. The WHOLE layout is replaced rather than patched, because a column order is only meaningful entire and a patch could not tell "I cleared my sort" from "I did not mention it". Written by the grid itself; delete_table-preferences_tableKey resets it to the workspace default. PUT /v1/table-preferences/:tableKey (access: workspace)

put_workspace_automated-emails

ChatGPT
Replace the workspace's automated emails Turn the workspace's system mails on or off and set the name and reply-to they go out under. Admin only, and it replaces the whole document rather than patching — send what the read answered with the changes merged in, or the mails you left out are the ones you turned off. A kind outside the vocabulary, an over-long subject prefix, or a reply-to that is not an email address is refused with the field named. PUT /v1/workspace/automated-emails (access: workspace:admin)

put_workspace_directory_provisioning

ChatGPT
Set directory provisioning Sets whether a directory sync invites the people it finds (autoInvite) and which seat those invites carry (defaultRole: admin, team_lead or member - never owner). Both fields are optional and an absent one is left alone. Reach for it to switch automatic provisioning on or off for a connected Google or Microsoft domain. PUT /v1/workspace/directory/provisioning (access: workspace:admin)

put_workspace_disclosure-states

ChatGPT
Replace the states this workspace adds Replaces the states this workspace ADDS to the statutory seven for commercial financing disclosures. The seven (CA, NY, UT, VA, GA, FL, CT) cannot be edited or removed — naming one is a 400 saying so — and an empty list leaves them in place. PUT /v1/workspace/disclosure-states (access: workspace:admin)

put_workspace_domains_id_form-host

ChatGPT
Serve forms from a workspace domain Names the address under one of the workspace's verified domains (apply.acmecapital.com) that its form links use instead of the default, and answers the domain with its formHost: the CNAME record to create and a pending status. One per workspace - naming another replaces it. Refused with 409 when the domain is not verified, 400 when the address is not under the domain (or is the domain itself, which DNS cannot point anywhere), 503 when this deployment serves no custom form addresses. Use the domain's recheck once the record is published; links switch over only when it is found. PUT /v1/workspace/domains/:id/form-host (access: workspace:admin)

put_workspace_execution-profile

ChatGPT
Save the execution profile Set or clear who signs for this workspace: a name, an officer title and an address, all three together or all three blank. Reach for this from Settings then Workspace. It is an admin act, because naming who binds the company is what its paper then claims. PUT /v1/workspace/execution-profile (access: workspace:admin)

put_workspace_heartbeat

ChatGPT
Update workspace automated follow-up settings Update the workspace's automated follow-up (heartbeat) settings, including whether it is enabled, its schedule, allowed actions and review preferences. Requires a workspace administrator. Supplied fields replace their saved values; omitted fields retain their existing values. The server validates combinations and returns the saved settings. These settings control subsequent background runs; this request does not immediately run the worker or send a follow-up. PUT /v1/workspace/heartbeat (access: workspace:admin)

put_workspace_heartbeat_me

ChatGPT
Elect yourself in or out of heartbeat follow-ups A rep's own switch: whether the heartbeat drafts follow-ups on the caller's own leads. Send {enabled}. It writes only the caller's election and answers the workspace's settings with the election list as it now stands. Only when the workspace runs follow-ups per rep (follow_up.per_rep); under one workspace-wide switch it is a 400 saying an admin turns that on first. Any seat with an account. PUT /v1/workspace/heartbeat/me (access: workspace)

put_workspace_heartbeat_stops_id

ChatGPT
Stop or resume the heartbeat for a conversation Send {"stopped": true} to make the heartbeat write nothing more for this conversation — no reply drafts, no follow-ups — and for every other conversation filed under the same record; the drafts already waiting for them are settled as refused and the count comes back as refused. {"stopped": false} resumes it. :id is the conversation id. Any seat that can read the conversation; nothing is sent either way. Use it when a person says to leave a lead, deal or merchant alone; to stop a whole status or board use the heartbeat settings instead. PUT /v1/workspace/heartbeat/stops/:id (access: workspace)

put_workspace_portal-access

ChatGPT
Set how long client portal access lasts Sets the workspace's portal access rule: {mode: while_open | link_expiry | days_after_close, days?: 1–365, only with days_after_close}. Takes effect on every borrower's next call; revoking an invitation still ends access at once. Admin seat; a 400 names a mode or days outside the rule. PUT /v1/workspace/portal-access (access: workspace:admin)

put_workspace_preferences_key

ChatGPT
Replace workspace preferences key Publish one shared default for the whole workspace as {"value": <json>}; JSON null clears it. The key must be namespaced ('table:leads', 'setting:theme') and an oversized value is refused. ADMIN ONLY, unlike the read beside it — a member setting their own override uses PUT /v1/preferences/:key. PUT /v1/workspace/preferences/:key (access: workspace:admin)

put_workspace_processor_criteria

ChatGPT
Change what counts as a processed deal Replace this workspace's processor@ criteria whole: intake, completeness, dealCreation (on_arrival or when_complete), emailsOff, dealStatus, defaultAssigneeAccountId. Strict — an unknown key, signal, check, document type, severity or status is a 400 naming it, and the default assignee must hold a seat. Admin only. PUT /v1/workspace/processor/criteria (access: workspace:admin)

put_workspace_processor_intake

ChatGPT
Choose where applications arrive Sets the workspace's intake address: empty for the platform's shared address, otherwise one of the workspace's own processor accounts or an alias of one. Anything else is refused. PUT /v1/workspace/processor/intake (access: workspace:admin)

put_workspace_processor_result-mappings

ChatGPT
Save where a file type's results go on a board Replaces the result mappings of one file type on one board, for one file-column scope: {documentType, definitionId, fileColumnId (null = any file on the record; a file column narrows the set to files held in it), mappings: [{field, target: same_row|linked_row, relationColumnId, destinationColumnId, updateRule: fill_empty|review|overwrite, createMissingLink}]}. The board's other scopes are left as they are; the answer is the board's whole set. Every column must be on the board (or, for linked_row, on the board the relation column points at) and able to hold a value — not a relation, person, file or computed column; one field per destination column. createMissingLink makes and links a record on the related board when the relation links nothing. [] clears the scope. Unknown keys are refused. Admin only. PUT /v1/workspace/processor/result-mappings (access: workspace:admin)

put_workspace_processor_submission-sender

ChatGPT
Choose the submission email Sets the workspace's submission email: one of its own connected accounts or an alias of one, which its own-lender submissions are sent from and whose replies are read. Empty clears it, which refuses own-lender sends until another is chosen. PUT /v1/workspace/processor/submission-sender (access: workspace:admin)

put_workspace_role-presets_presetId_members_accountId

ChatGPT
Give a member a role Applies a STAFF preset to one member in one step: their seat becomes the preset's, each board of the preset's pack gets the preset's level (a board it does not name goes back to the default), and their books grant follows the preset's level (none revokes it). Answers every level it wrote. An owner's seat is above every preset (409); an affiliate preset is never applied to a seat (400). A workspace admin's. PUT /v1/workspace/role-presets/:presetId/members/:accountId (access: workspace:admin)

put_workspace_sidebar-folders_id_items

ChatGPT
File items into a sidebar folder Replaces the folder's items with targets, in order: route words, or "board:<board id>". A target another folder held MOVES here (a row sits in one place); one this list leaves out returns to the rail; [] empties the folder. Items the caller cannot open are kept, after the ones sent, so a member rearranging a folder never drops a board they were not shown. Filing a board the caller cannot open is refused as not found. Any seat. PUT /v1/workspace/sidebar-folders/:id/items (access: workspace)

put_workspace_sidebar-folders_order

ChatGPT
Reorder the sidebar folders Sets the folders' order for the whole workspace. order lists EVERY folder id once, first to last; a list that leaves one out or names one twice is refused. Answers the folders as the read does. Any seat. PUT /v1/workspace/sidebar-folders/order (access: workspace)

put_workspace_super-kyc

ChatGPT
Replace the workspace's Super KYC settings Save the rules Super KYC judges this workspace's links by. The stored settings are the starting point, so a field left out is left alone; a list that is sent replaces the stored list whole. Amounts are in cents, $100 to $10,000,000; nsf_count is 1–50 and nsf_window_days 1–365; each list holds at most 100 entries of 80 characters. A lender name made only of generic words ("Capital Funding"), a name in both extra_lenders and ignored_lenders, or a watch term that is neither a name nor a domain is a 400 naming the field and the entry. It takes effect at each link's next check and leaves findings already filed alone. ADMIN ONLY, while reading is any seat's. PUT /v1/workspace/super-kyc (access: workspace:admin)

put_workspace_teams_teamId_members_accountId

ChatGPT
Put a member on a team Put one member on one team, answering the team as it now stands. The address names the pair, so adding somebody who is already on it is not an error. Admin only. PUT /v1/workspace/teams/:teamId/members/:accountId (access: workspace:admin)

put_workspace_underwriting-criteria

ChatGPT
Set the workspace's underwriting criteria Replace the scorecard. Every criterion sent is created or updated by its key, and EVERY ACTIVE CRITERION NOT SENT IS RETIRED — sending an empty list retires the whole template, which is a real request and not an accident the route will second-guess. At most 200 criteria; a blank key, or the same key twice, is refused rather than silently resolved. Category defaults to personal_credit and value kind to pass_fail. Admin only — the template is shape. PUT /v1/workspace/underwriting-criteria (access: workspace:admin)

put_workspace_underwriting_settings

ChatGPT
Save the underwriting settings Turn automatic re-evaluation on or off. On means every bank-statement, tax-return, credit-report or extraction delivery for a deal re-runs the published rules and can open a manual review. PUT /v1/workspace/underwriting/settings (access: workspace:admin)

recalculate_bank_statement

ChatGPT
Recalculates stored bank statement aggregates and risk signals for one visible statement. Use only when the user asks to refresh/recalculate statement analysis. This writes derived analysis fields and may affect underwriting views.

recalculate_tax_return

ChatGPT
Recalculates stored tax return risk/profile analysis for one visible tax return. Use only when the user asks to refresh/recalculate tax-return analysis. This writes derived analysis fields and may affect underwriting views.

refresh_specter_session

ChatGPT
Refreshes an existing Specter Systems session with a refresh token from the auth backend. Use only when app-level OAuth is unavailable and an authenticated user's access token has expired. It sends the refresh token in the Authorization header; model-visible output redacts new tokens and secrets.

reject_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Rejects a pending Heartbeat action so it is not applied. Use when the user declines the proposed next step. Optionally include a reason.

remove_assignment_for_entity

ChatGPT
For non-merchant Specter broker/operator users only. Removes ONE person or team from ONE role on ONE record, leaving everyone else in that role untouched. Prefer this over set_assignments_for_entity when the user named somebody to take off — it cannot accidentally unassign the others. The removed person is notified, so it is marked destructive.

reorder_fos_setup_deal_statuses

ChatGPT
Tenant-admin-only setup helper. Reorders deal/facility statuses for the authenticated tenant. Use only when an admin explicitly asks to change status ordering.

reprocess_facility_documents

ChatGPT
Re-runs document analysis over the files a facility ALREADY has and fills in extracted fields that are empty. It cannot overwrite data: if the merchant already holds contact details, owner name, SSN, date of birth or EIN, the call is REFUSED with 409 would_overwrite_existing_data and nothing is written. Use when a facility's documents were misread and its fields came out blank. It SPENDS processor and model budget on every call and is not free to retry, so name the facility and confirm with the user before calling, call it for one facility at a time, and never retry a failure without asking again. On a 409, report the populated_fields list from the response verbatim — those are the fields blocking the repair, and the user needs to know which to clear or fix in the app. Not available to merchant applicant accounts.

request_missing_docs

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL stip chase for a facility: the backend computes which documents are missing (bank-statement months with no coverage, expected tax years absent), composes a request message listing exactly those items (or uses your message override), and sends it to the merchant's channel. Returns the computed gap list plus per-step results. PREFER THIS over manually reading statements/tax returns and calling send_channel_message when the user asks what's missing on a file or to chase documents.

revert_heartbeat_action

ChatGPT
For non-merchant Specter broker/operator users only. Undoes an APPLIED Heartbeat action by restoring the fields it changed to their previous values — only if they have not changed since (otherwise it reports a conflict and skips). Only field updates are revertible; outbound messages cannot be undone. Use when the user wants to roll back a Heartbeat change.

search_records

ChatGPT
For non-merchant Specter broker/operator users only. Searches the tenant's CRM records — leads, merchants/counterparties, deals, and more — by ONE free-text query that matches name, business name, first/last name, email, or phone. This is the DEFAULT way to turn any person or business the user names into a concrete record with its id: use it FIRST for a lookup, a who-is / find / pull-up question, or before drafting, messaging, or acting on someone — the name, email, or phone the user gave IS the query. Do not ask the user for an id and do not hand-build a REST path when this fits. It resolves specific records, not whole-collection totals: for how-many counts, read the collection endpoint instead. Returns each match's entity_type, entity_id, name, and a short subtitle. Read-only; tenant-scoped and RBAC-enforced.

search_workflow_agents

ChatGPT
Searches/list agents in the authenticated tenant for Workflow V3 assignment configuration. Use to resolve a named user/agent or current profile user to a verified agent_id before configuring assignment actions or task assignee_id. It is read-only; never invent an agent_id if no exact or clearly intended match is returned.

search_workflow_builder_tools

ChatGPT
Read-only local reference search for Workflow V3 terminology. Returns suggested entity, trigger, field and node names and references to separately registered tools. It never executes, registers, unlocks or dispatches any other tool, never reads tenant records, and never creates, updates, validates or publishes a workflow. Every referenced operation is independently listed with its own input schema and annotations and must be called separately. This helper is optional.

select_specter_tenant

ChatGPT
Completes sign-in for users who have access to more than one tenant. Use after sign_in_to_specter returns requires_tenant_selection=true. It sends the tenant_selection_token and selected tenant_id to the live auth backend; model-visible output redacts session tokens and secrets.

send_blast_retarget

ChatGPT
For non-merchant Specter broker/operator users only. Re-sends an SMS blast to the slice of its original audience named by target — by default everyone who never replied. This sends real texts to real merchants and CANNOT be undone. There is NO recipient ceiling: the audience is whatever the preview resolves to, which can be thousands. So call preview_blast_retarget FIRST, tell the user the eligible_count, and pass that same number as confirm_recipient_count here; a mismatch means you are sending to an audience the user never saw. Pass the SAME target you previewed — the send recomputes the audience server-side, so a different target reaches a different set. Counts from the preview are advisory: state landed/failed from THIS call's response, never the preview's. Re-running an identical retarget is refused with a no-recipients error because each source recipient is stamped once, but changing target produces a fresh audience with no second confirmation, so treat every target change as a new send that needs its own preview and approval.

send_channel_message

ChatGPT
Sends a message in an existing visible communication channel as the authenticated user. channel_id is required and this tool does not create a missing channel. THE CHANNEL IS THE RECORD THE MESSAGE IS FILED UNDER, so it must be the channel of the person you are writing to — never reuse a channel you happen to have from an earlier step. Messaging somebody through an unrelated record files your message in that record's history and checks the wrong contact against do_not_contact. If the recipient has no record, say so and offer to create one rather than borrowing a channel. Resolve the exact ID with get_communication_channel_for_entity; never guess a communications REST path. Use only after the user explicitly asks to send or reply. For email/SMS, include only the intended recipient fields and message body; backend sender permissions, channel access, dispatch, and audit logging apply. A SUCCESSFUL RESPONSE MEANS QUEUED, NOT DELIVERED: the backend accepts the message, returns 200, and dispatches to the carrier afterwards, so a send that fails at the carrier (no connected sender, rate limit, expired auth, bad number) still answers 200 here. Report it as "queued" or "sending" and NEVER as "sent". To find out what actually happened, read the message back with list_channel_messages: status becomes sent or failed, and a failed row carries dispatch_error and failure_class saying why.

send_channel_messages

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL batch outreach: sends one personalised message to each of up to 25 recipients, resolving each recipient's channel for you. PREFER THIS over a chain of get_communication_channel_for_entity + send_channel_message calls whenever the user asks to message, text, or email MORE THAN ONE person — working through a list, a batch, or 'the next few'. One call means the user approves the batch once instead of once per recipient. Recipients are independent: an invalid number fails only its own entry, and the response reports every recipient's outcome with sent/failed counts, so report queued, skipped, and rejected entries separately. An ok entry is queued, not proof of carrier delivery. Duplicate destinations after the first entry are skipped. Include expected_last_message_id when replying to a conversation you have reviewed; if it changed, reread before composing a replacement. Never retry the whole batch after partial success or an ambiguous timeout. Each entry needs the message text plus recipient_phone (sms) or recipient_email (email) — supply the real contact you looked up; the backend will not substitute one. Identify the thread with channel_id when you already have it, otherwise entity_type + entity_id.

send_conversation_message

ChatGPT
Sends a reply in a visible conversation as the authenticated user. Use only after the user explicitly asks to send or reply. The backend fills the conversation recipient, verifies sender access, dispatches email/SMS where applicable, and records audit data. A SUCCESSFUL RESPONSE MEANS QUEUED, NOT DELIVERED: dispatch to the carrier happens after the 200, so a send that fails at the carrier still answers 200 here. Report it as "queued" or "sending", never "sent", and read the message back with list_conversation_messages to learn what happened — status becomes sent or failed, and a failed row carries dispatch_error and failure_class.

send_offer_to_merchant

ChatGPT
For non-merchant Specter broker/operator users only. ONE-CALL offer delivery: creates the financing offer on the facility, formats a summary (amount, factor rate, term, payback), and sends it to the merchant's channel — with per-step results. PREFER THIS over separate create_offer + send_channel_message calls when the user asks to send or present an offer. Steps stop on the first failure.

set_assignments_for_entity

ChatGPT
For non-merchant Specter broker/operator users only. Assigns people or teams to one role on one record — "assign this deal to Zach", "put Yoni and me on it", "make Sarah the closer". This REPLACES the role's entire assignee list. Anyone currently in that role and absent from assignments is unassigned and notified they were removed, so read get_assignments_for_entity first and send the existing people back alongside the new one whenever the user said ADD rather than "only". Send user_id from list_team_members, not id: the backend looks the number up in the users table first and only falls back to treating it as an agent id, so an agent id that collides with an unrelated user id silently assigns the wrong person. Roles are per entity type and a role the entity does not use writes a row nothing reads: facility -> opener, closer, originator, processor, assigned, admin, team_lead; counterparty -> assigned, admin; lead -> owner, admin; submission -> assigned, owner, sent_by; task -> assigned; bank_statement -> owner, closer, admin; tax_return -> owner, admin; business_research -> assigned, admin; credit_memo -> prepared_by, reviewed_by, approved_by; credit_report -> owner; capital_source -> admin; fos_call_record -> assigned; fos_sms_record -> assigned. A TEAM is not a role: assign a team under the entity's normal role with type=team. business_research carries assignments too, but only get_assignments_for_entity can reach them: the write routes resolve entity_type through the RBAC resource map, which has no case for it, and answer 400 before the handler runs. Assigning notifies the people involved, so it is marked destructive.

set_assignments_for_records

ChatGPT
For non-merchant Specter broker/operator users only. Puts the SAME people on the SAME role across many records at once — "assign all the missing-docs deals to me", "give Yoni the whole Salesfloor queue". One call, one role, one assignee list, many record ids. Every caveat on set_assignments_for_entity applies here multiplied by the number of records: this REPLACES the role on each one, so anyone currently holding it on any of those records and absent from assignments is unassigned — SILENTLY, because unlike the single-record write this path sends no removal notice, so the people dropped never find out. There is no per-record exception — if the records have different people on them today, assign them one at a time instead. Send user_id from list_team_members, not id. Only facilities and leads support bulk assignment — every other entity type is rejected outright, so assign those one at a time. Roles are per entity type and a role the entity does not use writes a row nothing reads: facility -> opener, closer, originator, processor, assigned, admin, team_lead; counterparty -> assigned, admin; lead -> owner, admin; submission -> assigned, owner, sent_by; task -> assigned; bank_statement -> owner, closer, admin; tax_return -> owner, admin; business_research -> assigned, admin; credit_memo -> prepared_by, reviewed_by, approved_by; credit_report -> owner; capital_source -> admin; fos_call_record -> assigned; fos_sms_record -> assigned. A TEAM is not a role: assign a team under the entity's normal role with type=team. business_research carries assignments too, but only get_assignments_for_entity can reach them: the write routes resolve entity_type through the RBAC resource map, which has no case for it, and answer 400 before the handler runs. Notifies everyone newly assigned, once per bulk call rather than once per record. It is marked destructive for the removals it does not announce.

set_lead_status

ChatGPT
For non-merchant Specter broker/operator users only. Moves a lead to a new pipeline status — the usual write after a call, text or reply. Transitions are VALIDATED against the lead's current status, so not every target is reachable from where the lead is now: read the current status with get_lead first, and if the change is rejected say which status the lead is in rather than retrying. Pass notes to record why it moved. This writes live CRM/FOS data.

show_application_summary

ChatGPT
Renders a compact application or intake summary card from already retrieved data. Use after a data tool when a concise visual summary helps the user. This render-only tool does not call external systems or mutate data.

sign_in_to_specter

ChatGPT
Authenticates an existing Specter Systems user with email and password. Use only when the user explicitly asks to sign in and app-level OAuth is unavailable. This calls the live auth backend and may return a tenant-selection challenge; model-visible output redacts session tokens and secrets. It does not create, update, or delete CRM/FOS business records.

submit_public_financing_application

ChatGPT
Submits a completed public financing application for a merchant or broker. Use only after the applicant reviews the answers and consents to submit them to Specter Systems. This creates or updates live intake/application records and may trigger downstream intake automation. Do not collect or submit SSNs, payment card data, health data, or unrelated personal data.

transition_application_stage

ChatGPT
Transitions an authenticated facility/application to a lifecycle stage accepted by the FOS workflow. Use only when the user explicitly asks to move an application stage. Backend lifecycle gates, tenant isolation, permission checks, notifications, and audit history apply.

update_bank_statement

ChatGPT
Updates narrow bank statement header fields visible to the authenticated user. Use only when the user explicitly asks to correct statement metadata such as its name, status, or Drive folder URL. Do not use to submit raw financial account credentials or card data.

update_bank_statement_period

ChatGPT
Updates narrow period-level bank statement analysis fields visible to the authenticated user. Use only when the user explicitly asks to correct extracted period metrics. Do not collect full account credentials, card data, or unrelated sensitive data.

update_facility

ChatGPT
Updates fields on one authenticated facility/application — most commonly its STATUS (a tenant-defined deal status slug such as closed_follow_up; list them with list_fos_setup_deal_statuses), plus notes, follow-up notes, priority, or assigned agent. Use this for status/group moves like "move to closed follow up" — those are statuses, NOT lifecycle stages (transition_application_stage only handles intake/underwriting/decision/funding/servicing/outcome). Backend tenant isolation, permission checks, and audit history apply.

update_fos_setup_deal_status

ChatGPT
Tenant-admin-only setup helper. Updates a deal/facility status definition by stable value in the authenticated tenant. List statuses first unless the exact value is already known from the setup API.

update_fos_setup_group

ChatGPT
Tenant-admin-only setup helper. Updates a manual group for Deals, Leads, or Merchants by ID, scoped to the authenticated tenant and entity type. List groups first unless the exact group_id is already known from the setup API.

update_lead

ChatGPT
For non-merchant Specter broker/operator users only. Edits a lead's contact and business details — the same fields a person edits in the lead record. Only the fields you pass are changed; omit everything else and it keeps its current value. Call get_lead first so you are not overwriting something someone entered. Two fields are deliberately not editable here: status is validated as a pipeline transition, so use set_lead_status; and the lead's free-text notes are never shown to you, so writing them here would replace text you cannot read — use add_note, which appends a separate note rather than overwriting the field. This writes live CRM/FOS data.

update_lender_submission

ChatGPT
Updates narrow fields on a lender submission visible to the authenticated user, such as lender response, contact, requested terms, offered terms, and approved terms. Use only after the user asks to change those submission details. Backend permissions, tenant isolation, validation, and audit/event handling still apply.

update_lender_submission_status

ChatGPT
Transitions an authenticated lender submission to a new workflow status. Use only when the user explicitly asks to update status. This may trigger notifications or downstream workflow behavior, so it is marked destructive.

update_tax_return

ChatGPT
Updates narrow tax return header fields visible to the authenticated user, such as name, status, folder URL, and profitability summary. Use only when the user explicitly asks to correct tax return metadata.

update_tax_return_year

ChatGPT
Updates narrow year-level tax return analysis fields visible to the authenticated user. Use only when the user explicitly asks to correct extracted tax-year metrics. Do not collect or submit SSNs through this tool.

update_workflow_draft

ChatGPT
Updates an existing Workflow V3 definition for the authenticated tenant. Use when the user approves changes to a draft or wants to repair validation errors. Validate the updated graph before publish. Backend tenant permissions and validation still apply.

upload_application_document

ChatGPT
For non-merchant Specter broker/operator users only. Merchant applicant accounts must not use this tool; they are limited by backend enforcement to signed-form submission, public document upload, and merchant-portal/applicant-facing updates. Uploads a ChatGPT-supplied bank statement or supporting document to an authenticated Specter facility/application. Use immediately after create_application_facility when the applicant attached files in chat. This stores the file in Specter, attaches it to the facility files list, creates a document link, and starts document preprocessing when possible. Do not say files are attached unless this tool succeeds for the supplied files.

upload_public_bank_statement

ChatGPT
Uploads a bank statement or supporting document to a specific public signed-form submission field. Use only when the applicant has supplied a file for their application. This writes the file to live Specter storage and can replace the file association for that form field depending on backend rules.

validate_public_financing_application

ChatGPT
Validates a public merchant or broker application before final submission. Use after collecting answers from the signed form. The applicant_data object must contain only fields from the fetched form schema; do not request SSNs, card data, health data, or unrelated personal data.

validate_workflow_draft

ChatGPT
Validates a proposed Workflow V3 graph without saving or publishing it. Use after drafting or modifying a workflow and before create, update, publish, or execution. Inputs must use registered triggers, node types, handles, and config schemas from the workflow catalog.

Capabilities

WritesInteractive

Example Prompts

Click any prompt to copy it.

Similar Apps