MCP App Store
by KeenEthics
Entendre icon

Entendre MCP Server for Claude and ChatGPT

By Entendre AI

Overview

Connect Claude to Entendre’s Unified API and work across your accounting data from a single conversation—browse accounts, query transactions, review balances, classify activity, create journal entries, generate reports, and keep your general ledger in sync without exporting data or switching tools.

How to connect

  1. Claude: Open the Entendre listing in the Claude connectors directory and choose Connect. Claude listing Entendre exposes 3 tools here.
  2. ChatGPT: Open the Entendre listing in the ChatGPT app directory and choose Connect. ChatGPT listing Entendre exposes 116 tools here.

Tools

apply_cash_application

ChatGPT
Apply a draft cash-application Applies an existing draft cash application in one of two modes. With compiled_rules it posts the journal entry, claims the deposit and, depending on settings, writes to Stripe to mark the invoice paid. With journal_entry_id it links a journal entry you already posted, after checking that it is posted, balanced, touches the bank account, is not linked elsewhere and has the customer tag. The compiled rules represent verified active policy; [] represents a verified policy without overrides. The applied match and its Stripe state.

cancel_agent_run

ChatGPT
Cancel an agent run Cancels an agent run by sending a stop request. It needs write:agents and a user-owned key or a user session; an ownerless key gets 403. meta.cancellation_status says what happened: accepted (stop signal confirmed), pending (not confirmed), no_active_turn or no_session (nothing was running), owner_mismatch (ownership not verified). The request's acceptance is not completion: check the run later with list_agent_runs. Work already done is not undone. The cancellation status.

cancel_request

ChatGPT
Cancel request Cancels an integration request and invalidates its client links. Needs firm write access. Sources the client already connected through the request stay connected. Cancellation outcome.

classify_transactions

ChatGPT
Classify transactions and post eligible entries asynchronously Classifies the selected transactions with the organization's rules and posts POSTED journal entries at once for confident matches that have a category and a payment account; the rest stay pending_review or unclassified. Select by sequence_numbers, transaction_ids, or a date range (at most 100 transactions). Auto-posting cannot be turned off, and the entries are not synced to the GL. It returns a cls_ job: read the result with get_classification in a later call. With search_term, send unaccounted_only: false. A classification job id; confident matches post.

close_accounting_period

ChatGPT
Close an accounting period Closes an accounting period: it records closing balances, locks posted journal entries and carries the balances into the next open period. It fails when an earlier period is still open or no next period exists. Undoing it needs an admin reopen, which can change later balances. The closed period.

complete_cash_application

ChatGPT
Complete an applied cash-application Completes an applied cash application (applied, stripe_status: pending) by marking its Stripe invoices paid; the row becomes completed. It is safe to repeat on a completed row. With require_gl_sync: true, the call returns 409 unless the linked journal entry is posted and synced to the GL. Without it, only organizations on Campfire get that check, so Stripe can show the invoice paid while the GL has no entry. It returns 400 when the row is not applied, 409 on a lock, an unpaid invoice or a disconnected Stripe connection, and 502 on a Stripe error. The completed match, or the lock, state or provider error.

create_accruals

ChatGPT
Create vendor accruals Creates vendor accruals in one batch. Each row succeeds or fails on its own: the response lists created and failed rows, and each failure says why (for example a duplicate for that vendor and period, an unmapped vendor account, or a closed period). Amounts must be above zero. There is no account field: each accrual books to the vendor's mapped expense account (default_expense_account_id in vault.list_vault_vendors); a QuickBooks vendor without one uses the account of its recent bills, else the row fails. An accrual here does not show whether the vendor's bill already exists in the GL. created and failed rows.

create_and_apply_cash_application

ChatGPT
Create and apply a cash-application in one call Creates and applies a cash application in one call: it posts the journal entry and claims the deposit; unapply_cash_application reverses it only before the Stripe sync. Each allocation must equal the invoice's amount_due (no partial payments), and all invoices must belong to one customer. The deposit rules and 400 reason values are the same as for preview_cash_application. The compiled rules represent the organization's verified active policy; [] represents a verified policy without overrides. Automatic application requires verified rules. The applied match with its journal entry.

create_asset

ChatGPT
Create a manual asset Creates a manual asset lot for an off-chain or manually tracked position, and adjusts the matching ledger account balance. Existing lots are available through list_assets. To correct an existing manual lot, use update_asset. The created lot and its balance adjustment.

create_client

ChatGPT
Create client Creates a draft client and its organization; omit name and web_address to generate both. Needs firm write access, a firm administrator; client-restricted keys cannot call it. Duplicate-name comparison ignores surrounding whitespace, case and repeated spaces; similar names may identify an existing client. A failed roster read does not prove the client is absent. Creation does not activate the client. An idempotency_key is required for this request. Reuse it only when retrying the same request. The new draft client and its organization.

create_journal_entry

ChatGPT
Create a journal entry Creates a journal entry in one of two ways. From a transaction: send transaction_id, a postable ledger_account_id and transaction_type; the lines come from the transaction's gross fiat value, direction and source, and a non-stablecoin crypto outflow books 4 lines with cost basis and realized gain or loss (cost basis must be computed first). Manual: send balanced lines, a status of draft or posted, and at most one transaction_id. A transaction entry is DRAFT unless you send auto_post: true. If the target account has no GL mapping (list_account_mappings), the entry cannot reach the GL. The new entry with its status.

create_ledger_account

ChatGPT
Create a ledger account Creates a ledger account with a name and a type (Asset, Liability, Equity, Income or Expense). sequence_number defaults to the next number in that type's block, and parent_account_id puts the account under a parent. Ledger accounts belong to the organization, not to a legal entity: leave legal_entity_id out and do not ask the user for an entity. The new account.

create_request

ChatGPT
Create an integration request for a client and send its email immediately. Supports wallets, exchanges, banks, QuickBooks, Xero, NetSuite, Ramp, Stripe AR and Gmail. Document requests are unavailable through this connector. Requires firm write access and an idempotency_key. The email cannot be undone. email_sent reports provider acceptance, not delivery. Returns the saved request and email send result. Do not retry an uncertain write automatically.

create_revaluation

ChatGPT
Trigger asset revaluation Starts the requested revaluation and returns acceptance, not completion. With period_id, the period must be open, earlier periods closed and the next period present and open; existing revaluation or reversal entries conflict. An already running period job is returned. Poll list_revaluations for this variant. With legal_entity_id instead, the existing entity-wide workflow returns a job_ configuration receipt, which is not in that list. This connector has no job-configuration polling tool. A repeat write is not a status check. The in-flight guard applies to the selected scope. The accepted period job or entity-wide job configuration receipt (202), or an existing period job.

create_source

ChatGPT
Create a source Creates a wallet or a Niural supplier source (201). For exchange, bank_account or card, it selects the source of an active connection with connection_id (200, nothing created) or starts a statement import with document_id (202, queued). Creating a wallet does not import its transactions. A 409 on a wallet means the address already exists on that chain: the existing source is available through list_sources. 201 created, 200 existing connected source, or 202 queued import.

create_tag

ChatGPT
Create a tag Creates a tag from a key in the closed set and a value label. The key and value pair is unique in the organization. Tags are labels for filters and reports; they change no balance. The new tag.

create_transaction

ChatGPT
Create transaction Creates one manual fiat transaction on an existing bank, card or other fiat source; it records an event and moves no money. Crypto sources are rejected. currency must be the legal entity's reporting currency, amounts are unsigned (direction sets the sign), and a timestamp in a closed period returns 422. It does not classify the transaction. The new transaction; its OT number can arrive later.

delete_agent

ChatGPT
Delete an agent Deletes an agent: it permanently removes the agent and all its run history, and stops its schedule at once. This cannot be undone. Pause the agent in Entendre to retain its history. Success with no body.

delete_asset

ChatGPT
Delete an asset Deletes an asset lot as a soft delete and reverses its balance adjustment in the same operation. The lot then leaves lists and future cost-basis calculations; only an operator can restore it. Lot details are available through list_assets. Success with no body; the lot is soft-deleted and its balance adjustment reversed.

delete_journal_entry

ChatGPT
Delete a journal entry Soft-deletes one DRAFT, ERROR or UNPOSTED journal entry that is not synced to a GL; it leaves every list and only an operator can restore it. Posted or synced entries cannot be deleted directly. Unposting is a separate operation and is blocked by a closed period. Success with no body, or an error when the status or a GL sync blocks the delete.

delete_ledger_account

ChatGPT
Delete a ledger account Soft-deletes a ledger account and its direct child accounts. It returns 409 when the account or any account below it is used by a journal entry line, asset, template, wallet or exchange source; balances are not checked. Accounts below the direct children stay active, so delete a deeper tree from the bottom up. Success with no body, or 409 when a reference blocks the delete.

delete_memory

ChatGPT
Soft-delete a memory document Deletes one memory document (path and reason required) or one operational rule (rule_id required, from list_memory). A document is soft-deleted to the trash with the reason, and purge: true deletes it permanently; this tool set has no restore. A rule delete stops the rule for later runs and does not undo past classifications or journal entries; an unknown id returns 404. Deletion is a separate requested action. The document or rule deletion outcome.

delete_source

ChatGPT
Delete one source Deletes one source through its type's own checks: a wallet or exchange used by journal entries, a bank account with accounted transactions, or a Niural source with non-draft invoices returns 409. A Rain or Ramp card or a Ramp bank account deletes with its transactions; one whose transactions have journal entries returns 409. Staking sources cannot be deleted (422); the user disconnects the provider instead. confirm_source_type must match the source's real type, and the call needs read access to sources as well as write access. Deleting a Plaid account also soft-deletes its bank transactions, and deleting the last one removes the Plaid item. Success, or 409 or 422 when a guard blocks the delete.

delete_tag

ChatGPT
Delete a tag Soft-deletes a tag that is not in use, so it no longer appears in lists. A tag still assigned to journal-entry lines or templates returns 409. Deletion does not unassign references. Success with no body, or 409 when the tag is in use.

delete_transaction

ChatGPT
Delete a transaction Soft-deletes one transaction. A transaction booked by a posted journal entry returns a conflict; unposting is a separate operation that requires an explicit user request. A draft entry can stay in place after its transaction is deleted. Success with no body, or a conflict when a posted entry books the transaction.

delete_vault_document

ChatGPT
Delete a document Soft-deletes one Vault document, so it no longer appears in lists; bill, receipt and statement rows that use it stay. Deletion is a separate requested action. A second delete of the same id returns 404. Success with no body.

disconnect_connection

ChatGPT
Disconnect connection Disconnects one connection through its provider's cleanup. Plaid cleanup removes linked bank accounts and transactions while retaining journals, and accounting records can block it; other providers keep different resources. The result contains cleanup, retained_resources and revocation_error. Incomplete cleanup is not a complete disconnect. Cleanup results, retained resources and any revocation error.

get_accounting_period_close_preflight

ChatGPT
Pre-close audit for an accounting period Returns the close blockers for one period: structural conditions, unbooked transactions and draft journal entries. It does not close the period. safe_to_close means only these checks passed; unposted, in-progress or failed entries, sync failures, clearing, revaluation and asset checks are not included. Preview arrays hold at most 25 rows; counts cover all matches. Blockers, recommendation, counts and short previews.

get_asset_tax_lots

ChatGPT
Get asset tax lots Returns acquisition lots as of a required as_of date: quantity, remaining quantity, cost basis, method and linked dispositions, oldest acquisition first. Disposals after as_of are added back. asset_types filters assets on the server. A refused or timed-out scan is incomplete. limit=25 — page with next_cursor Lot rows oldest first, and context with as_of and the method.

get_balance_sheet_report

ChatGPT
Balance sheet Returns Entendre book balances for Asset, Liability and Equity accounts in the selected periods. data_quality includes negative_asset_periods, unbalanced_periods and warnings that qualify the returned balances. Opening balances carry forward only from closed periods, so where earlier periods are open a column is period movement, not a cumulative position. Cash membership is not established by matching account names; cash and sign checks include both asset and liability types. For totals only, send include_zero_rows: false. No period named → send start_date and end_date for the last 3 UTC calendar months The 3 months include the current one: 1st of the UTC month 2 months back → end of this UTC month Account rows with period balances, ledger_accounts for the hierarchy, and data_quality.

get_classification

ChatGPT
Get classification status Returns a classification job: status, totals and one paginated items list in which each row's status says whether it was classified, pending review or unclassified. posting_status and journal_entry_id are set only on rows that an auto-post touched. succeeded means every row classified and posted; otherwise the job is partially_succeeded (a skipped post alone does not count). Status, items and totals.

get_closing_positions_report

ChatGPT
Closing positions Returns the accounting position of each token from the cost-basis lot subledger: quantity, cost basis, market value and unrealized gain, fiat included. It is not current live crypto exposure; for live holdings use get_treasury_balances. Money fields are 2-decimal amounts in one entity's reporting currency, which the response does not name, base_currency is available through list_legal_entities. data_quality is at the top level (not inside data) and covers every matched row; has_negative_positions identifies a limitation on portfolio totals. Prices use floating-point arithmetic and are not exact many-decimal values. Position rows, total_count across pages, and top-level data_quality.

get_cost_basis_settings

ChatGPT
Get cost basis settings Reads the connected organization cost-basis method and lot-tracking scope. Check configured before treating default values as a saved setting. For firm connections, select the client organization first. The methodology, accounting_based and configured flag inside data.

get_current_organization

ChatGPT
Get current organization Returns the organization's name, org_ id, web address, timezone and time format. There is no organization currency: each legal entity has its own base_currency (see list_legal_entities), so organization amounts have no single shared currency. No cost-basis policy is exposed here. Organization identity and display settings.

get_financial_insights

ChatGPT
Financial insights Returns the requested metrics for a window: revenue, profit, expenses, spending, burn_rate, runway, cash_flow, treasury_flow, each with value, formula, unit and details. basis is required: transactions is cash-basis and needs no posted entries; ledger is the posted-entry P&L by whole month. runway counts fiat and stablecoins only and is null with reason: cash_flow_positive when there is no burn. With basis: transactions, cash_flow covers only P&L-classified transactions, and details.data_quality lists what was excluded. For money in and money out of the bank accounts, wallets and exchange accounts, use treasury_flow with basis: transactions. State the basis in the answer. No dates named → send the trailing year ending today No basis named → basis=transactions One result per requested metric with its own details and quality notes; a missing value is not zero.

get_firm_settings

ChatGPT
Get firm settings Returns the firm name and its Entendre subdomain slug. Needs read:firms. The slug is not the company's website. The firm name and subdomain slug.

get_income_statement_report

ChatGPT
Income statement Returns Entendre income and expense balances by ledger account for the selected periods (accrual basis). This report provides accrual-basis revenue, expenses and net income, including when some transactions remain unposted. A populated report is not proof that the books are complete or reconciled, and an empty one is not proof of zero activity. Parent rows already include their children; adding every row double-counts balances. For totals only, send include_zero_rows: false. No period named → send start_date and end_date for the last 3 UTC calendar months The 3 months include the current one: 1st of the UTC month 2 months back → end of this UTC month Income and expense account balances, ledger_accounts for the hierarchy, and data_quality.

get_realized_gains_report

ChatGPT
Get realized gains report Returns sold asset lots with their disposals for the chosen accounting_period_ids; with no period it covers all history. With summary: true, limit: 1, summary.total_gain equals the sum of every page. When summary.by_currency has more than one entry there are separate currency totals and no flat totals. Gains use a per-unit estimate (sale_price × quantity_sold − cost_basis × quantity_sold), not a FIFO, LIFO, HIFO or average-cost calculation. Disposed lot rows; summary: true adds per-currency totals.

get_schedule_of_dispositions

ChatGPT
Schedule of dispositions Returns asset lots that have disposals, with each disposal's date, proceeds, cost basis and realized gain or loss. Each lot lists its full disposal history. For totals send summary: true; it gives the same per-currency summary as get_realized_gains_report over the same filters. Disposed lot rows; summary: true adds the per-currency summary.

get_treasury_balances

ChatGPT
Consolidated live treasury balances Returns the organization's live treasury position: one row per internal wallet, exchange account or bank account, with its tokens nested and total_fiat per row. It is the answer for "how much cash do we have now"; book balances come from get_balance_sheet_report. Wallets and banks come from the last stored refresh (as_of per row); exchanges from the daily snapshot. as_of (YYYY-MM-DD) gives a past position: banks are left out and a missing exchange snapshot is a gap (reason: no_snapshot). For a named month, use its last day as as_of. Every connected source has a row, "0.00" when empty. A failed source has total_fiat: null; check context.total_fiat_status before calling the total complete. For current crypto exposure, add up the crypto and exchange token rows by asset and keep bank cash separate. One wallet can hold two rows of one symbol (two contracts), so add every row, not only the first match. Source rows sorted by total_fiat, and context with warnings and total_fiat.

get_trial_balance_report

ChatGPT
Trial balance Returns debits, credits and the stored opening, closing and current balance for every account type over the periods in start_date–end_date (both required; a window with no periods returns 400). For one closed period, closing_balance is its opening balance plus its activity; it is 0 while the period is open. For more than one period, each balance field is the sum across the periods, not a position as of end_date. data_quality warnings qualify the balances; returned balances are not proof of reconciliation. An empty result means no balances in this window, not proof that the ledger recorded no activity. An end-only request uses the default window ending on that date. as_of: true selects explicit inception-to-date mode. Cash membership is not established by matching account names; cash and sign checks include both asset and liability types. The ledger_accounts array contains filtered accounts, including zero ancestors of retained accounts. meta.account_count is the filtered count and meta.total_account_count is the whole chart count. Their difference is accounts entirely zero for this period, not missing accounts. meta, with zero_rows_omitted, omitted_account_fields and view, comes back when you send include_zero_rows: false or verbosity: compact; it is absent only on the full faithful payload, where ledger_accounts is the whole chart. No dates named → send start_date and end_date for the last 3 UTC calendar months The 3 months include the current one: 1st of the UTC month 2 months back → end of this UTC month Account rows, ledger_accounts, data_quality, and meta with the returned and total account counts.

import_accounting_data

ChatGPT
Import accounting data Imports legal entities, the chart of accounts and customer and supplier tags from a connected accounting system; it imports no journals or transactions. Progress is available through list_imports. Multiple provider realms require an explicit selection. Bank, card and exchange connection ids return 404. The accepted import.

import_source_transactions

ChatGPT
Import a source's transactions Imports transactions for one source in the background. A 202 queued response has no status check and does not establish completion. A 200 skipped response means nothing was queued and includes the reason. Omit start_date for a refresh from the last cursor; send it for a dated backfill (wallets only, exchanges return 422). Bank accounts import through Plaid on their own (422). 202 queued, or 200 skipped with a reason.

invite_team_member

ChatGPT
Invite team member Invites a person to the firm and sends the invitation email; the send cannot be undone and the link expires after 14 days. Needs firm write access, a firm administrator. email_sent records whether the email provider accepted the send; a saved invitation does not prove the email went out. An idempotency_key is required for this request. Reuse it only when retrying the same request. The invitation and its delivery result.

list_account_mappings

ChatGPT
List account mappings Lists which Entendre ledger account maps to which GL account on one accounting connection, 25 a page at most. A row count is not a count of mapped accounts: for coverage, read every page and count distinct non-empty internal_ledger_account_id values against the same chart. Bank, card and exchange connection ids return 404. Mapping rows and a count.

list_accounting_periods

ChatGPT
List accounting periods Lists accounting periods, newest first, with status (open, soft_closed, closed) and close details. Periods are created as transactions are imported. Some organizations have several periods for one month (one per legal entity, plus archived ones): legal entity, start date and last update distinguish them. Name each entity with the period's legal_entity_name; do not match entities by list order. Period rows, newest first.

list_accruals

ChatGPT
List vendor accruals Lists the organization's vendor accruals whose journal entry still exists. Filter by vendor_id, period (YYYY-MM), status or journal_sequence_number. status defaults to posted; all returns every status but still leaves out accruals whose journal entry was deleted. Each call marks those orphaned accruals as reversed. Marks posted accruals whose journal entry was deleted as reversed. Every match in one response unless you ask for pages.

list_agent_runs

ChatGPT
List agent runs Lists one agent's runs, newest first, with results and errors. run_ids and status filter before paging; keep the same filters across pages. The cursor anchors to the last run, so new runs do not shift later pages. Run rows with results and errors.

list_agents

ChatGPT
List agents Returns every agent in one response, by name: schedule, triggers, capabilities, is_active and the latest run. agent_ids is the only filter. An agent is event-driven when it has an active *_webhook trigger, even with an empty schedule; if no trigger is active, nothing runs it. Schedule times are in the organization's timezone: get it from get_current_organization and name it. All agents with triggers, schedule and latest run.

list_assets

ChatGPT
List assets Lists digital-asset lots with quantity, cost basis and acquisition details. Filter by asset type, legal entity, ledger account, chain or date received. This is the accounting lot inventory, not live wallet balances. Lots with acquisition details.

list_cash_application_invoices

ChatGPT
List invoices eligible for cash application Lists locally stored Stripe invoices that a payment can match, filtered by customer_name (fuzzy), stripe_customer_id, status, amount with tolerance, and dates. status defaults to open. eligible is true only for an open invoice with an amount due; otherwise ineligibility_reason says why. When customer_name matches no customer or several, the page is empty and resolution gives the message and candidates: stripe_customer_id resolves the ambiguity. Invoice rows with amount due and eligibility, or resolution when the customer is unclear.

list_cash_applications

ChatGPT
List cash-applications Lists cash applications (payment-to-invoice matches). Filter by status (draft, applied, partially_applied, voided) and stripe_status (pending, synced, failed); applied and partially_applied need cursor paging, and an empty page can still have has_more: true. status_counts uses the older status names, not these filter values, and is computed independently of the status, stripe_status and cursor filters. Because this tool exposes no date parameters, {} here means the organization has none; the app scopes the same rollup to a deposit window, so its counts can differ. Draft matches are only part of unmatched deposits: zero drafts does not establish zero unmatched deposits. Match rows and the organization-wide status_counts.

list_clients

ChatGPT
List clients Lists the firm's client relationships with their organization ids and onboarding state. Filter by client id, status or name. Needs read:firms. A client relationship id (fcl_) is not the organization id. Clients with organization references and, when requested, filtered counts.

list_connections

ChatGPT
List connections Lists the organization's connections for GL, exchange, bank and card providers, with status, legal entities, source ids, capabilities and an ETag; credentials are never returned. Use a GL connection id with the tools.gl_mapping tools. Which families you see depends on the key's scopes (read:gl for GL, read:connections for the rest). Active authorization does not prove healthy data delivery. Sync and import outcomes are separate evidence. Listing can create missing connection records and reconcile pending authorization attempts from existing provider state. Connection rows with status, capabilities and source ids.

list_entity_mappings

ChatGPT
List entity mappings Lists which Entendre legal entity maps to which GL entity on one accounting connection. Get the connection id from list_connections. Bank, exchange and Rain card connection ids return 404. Entity mapping rows.

list_external_accounts

ChatGPT
List external accounts Lists the GL accounts imported from one accounting connection (QuickBooks, Xero, NetSuite, DualEntry or Campfire), 25 a page at most. This is the stored copy from the last import, not a live GL read, so an empty list does not prove the GL has no accounts. Bank, card and exchange connection ids return 404. External account rows and a count.

list_external_entities

ChatGPT
List external entities Lists the GL entities imported from one accounting connection. This is the stored copy from the last import, not a live GL read, so an empty list does not prove the GL has no entities. Bank, exchange and Rain card connection ids return 404. For a Ramp connection, the list is read live from Ramp on each call. External entity rows.

list_imports

ChatGPT
Get the connection's current import status Returns the latest import attempt of one accounting connection as a list of zero or one item; there is no history of runs. Several provider configurations return 409. Bank, card and exchange connection ids return 404. Zero or one import attempt.

list_integrations

ChatGPT
List integrations Lists every supported provider with its connection status: accounting (QuickBooks, Xero, NetSuite, DualEntry, Campfire, Intuit vendor login), cards (Ramp, Rain), receivables (Stripe, Request Finance), banking (Plaid), payroll (Finch), custody (Fireblocks) and communication (Gmail, Slack). Each provider has connected, count, unavailable and its connections with status, legal entity, external id (the realm for a GL) and last_activity_at. connected is true when a connection is active; unavailable: true means the status could not be read, not disconnected. GL, Ramp, Intuit, Gmail and Slack connections have no legal entity, so a legal_entity_id filter leaves them out; an unknown legal_entity_id returns 404. Status is a stored record, not a live check. One entry per provider with its connections.

list_invitations

ChatGPT
List invitations Lists the firm's pending and expired invitations. Accepted and revoked invitations are not returned; accepted members come from list_team_members. Needs read:firms. Invitation records.

list_journal_entries

ChatGPT
List journal entries Lists Entendre journal entries with their lines, newest accounting_date first. These are Entendre records, not a live read of the connected GL. Look entries up in batches by journal_sequence_numbers (JE-123) or transaction_sequence_numbers (OT-123); journal_entry_ids takes only je_ ids. Filter by statuses, sync_status (pending, synced, failed) with sync_error_types, accounting_date range, accounts, legal entities or tags. JE numbers are given at creation, so order by accounting_date, not by number. include_count: true with limit: 1 provides the count for scoping an unfiltered request. waiting / pending to sync → sync_status=pending; failed to sync → sync_status=failed duplicate sync → sync_error_types=DUPLICATE_DETECTED Default order: sort_by=accounting_date, sort_direction=desc, include_count=true Entries with lines, status and latest sync attempt; total_count counts the whole filtered set.

list_journal_sync_jobs

ChatGPT
List sync history records Lists past GL sync runs. By default only COMPLETED runs are rows; send include_all: true for failed, canceled and hung runs, or include_in_progress: true for running ones. status_counts counts every run status, so a nonzero failed count means the default page hides failures. UNRESOLVED counts runs whose status is unknown, not successes. This is Entendre sync history, not a live GL read. Run rows and status_counts.

list_ledger_accounts

ChatGPT
List ledger accounts Lists the chart of accounts in sequence_number order (sortable by sequence or name only). Accounts nest through parent_account_id; is_leaf: false marks a parent account, which can still hold its own postings. Archived accounts are hidden unless you send include_archived: true, while the app's chart shows them. With filters, include_count counts the filtered set, not the whole chart. Charts can contain unused default accounts; get_trial_balance_report provides activity evidence. Parent accounts have is_leaf=false; they can receive postings. Account rows with type, parent, leaf flag and normal balance.

list_memory

ChatGPT
List a memory directory Lists memory documents under a directory (kind=document, the default) as a frontmatter manifest without bodies or versions, or the operational rules (kind=rule), optionally for one domain. Rule listings reject cursor, limit and directory. A manifest does not contain policy text: to answer a policy question, search with search_memory and read the matches. An empty rule list does not mean the organization has no policy. To write a document back, read it with read_memory first to get its version. A document manifest, or the complete rule list with counts by domain when no domain filter is set.

list_organization_members

ChatGPT
List organization members Lists organization members, active and invited, with role and invitation status. Emails are masked by default (for example o***@entendre.finance); send include_pii: true only when the task needs full addresses. Inviting members and changing roles are not available here; the user does them in organization settings. Member rows with role, status and masked email.

list_organization_purpose_accounts

ChatGPT
List purpose accounts Returns the ledger account assigned to each accounting purpose for the organization. Each purpose has one row. ledger_account_id is null when no account is assigned. Requires read:organizations. For firm connections, select the client organization first. One row per purpose inside data, with the account ID, name and sequence number. Unassigned purposes have null account fields.

list_requests

ChatGPT
List requests Lists the firm's integration requests with their recipients, requested items, progress and submitted resource links. Filter by request id or status. Needs read:firms. A sent request does not prove the client connected the provider. Requests with recipients, items and submitted resource references.

list_revaluations

ChatGPT
List asset revaluation jobs Lists period-scoped revaluation attempts, newest first. Use it to poll only the period_id variant of create_revaluation. It does not contain the entity-wide job_ configuration receipt. Filter by period_id or status, and use fresh evidence for the selected attempt; an older completed attempt does not prove the latest one completed. Jobs with their status.

list_sources

ChatGPT
List sources Lists every source in one feed: wallets, exchanges, staking, banks (manual and Plaid), Ramp and Rain cards, Ramp bank accounts, Niural and Request Finance. Each row has common fields plus a detail object that varies by source_type, and every type has detail.ledger_account_id. Name lookup can return multiple matches; an empty match does not prove the provider is disconnected. On an unfiltered first page total_source_count is the organization total; include_count: true counts the current filter. A filter a source type cannot express (for example chains) drops that type; group_id filters by folder. Source rows with type, legal entity, status and detail.

list_tags

ChatGPT
List tags Lists the organization's tags (key and value labels), newest first. Keys come from a closed set: Customer, Supplier, System, ID, Bank Account, Cost Center, Class, Staff, Product, Workflow, Review Status, Custom. Filter with key, search or status. A filtered result with zero rows is complete for that filter. Tag rows with key, value and status.

list_team_members

ChatGPT
List team members Lists the firm's accepted members and their roles. Pending invitations come from list_invitations. Needs read:firms and a key that is not client-restricted. Accepted memberships.

list_transactions

ChatGPT
List transactions Lists transactions from connected wallets, exchanges, banks and cards, newest first. It is the main read for spend, vendor and amount questions; a transaction describes the economic event, and its accounting treatment is on the linked journal entry. has_posted_journal_entry: false means no posted entry exists, although a draft or error entry can. external.value_fiat.gross is the fiat value used on journal-entry lines; external.quantity is a token quantity, never dollars. Spam is excluded by default: include_spam: true adds spam rows and status: 'spam' lists only spam. The organization's skip list is not applied automatically. unclassified / uncategorized / unaccounted / not booked → has_posted_journal_entry=false excluded / skipped → the skip list in the server guidance not a filter spam → excluded by default; include_spam=true adds spam rows; status=spam lists only spam Transaction rows; total_count counts the whole filtered set, not the page.

list_vault_documents

ChatGPT
List Vault documents Lists Vault documents: bills, receipts, bank statements and other uploads. Send expand (bill, receipt, vendor, statement) to add the linked records as fields of each document; expand: 'statement' gives a statement's bnk_ id before an import. Rows are ordered by id, not by date. Timestamps identify document dates. An error is not an empty result. Requesting the download expansion records a document-view event in its history. Document rows, with the linked records you asked to expand.

list_vault_vendors

ChatGPT
List Vault vendors Lists Vault vendors with files_count and total_bills_amount, and default_expense_account_id when a vendor is mapped (filter with mapping_status). Use it to turn a vendor name into its vnd_ id with vendor_name_regex. Both totals cover all history and there is no date filter: total_bills_amount is billed payables, not spend for a period and not recognized expense. For vendor spend in a period use list_transactions with search and dates; for expense use get_financial_insights with basis: 'ledger'. An error is not an empty result. Vendor rows with file count, all-history billed amount and mapping.

patch_source

ChatGPT
Update one source Updates one source. Set ledger_account_id to assign an account or null to clear it; set group_id to move the source into a folder or null to take it out (staking and Request Finance sources cannot be in a folder). On a Rain card, send a non-null legal_entity_id only with an optional ledger_account_id: this moves the card and its transactions without journal entries to that legal entity, and resets their prices when the currency changes. Here a null ledger_account_id returns 400. Omitting it preserves the current assignment. Wallet fields (wallet_type, address, addresses, consolidate_fees, track_solana_rewards, status: pending_setup) go only with name, status, legal_entity_id, ledger_account_id, allow_nft_import and tag_ids; Niural fields (wallet_type, chain, address, name, provider_tags, a non-null legal_entity_id) go together and with a non-null group_id; other fields return 400 and change nothing. Omitted fields stay the same; fields the source type does not support return 422. The full updated source.

post_journal_entry

ChatGPT
Post a journal entry Posts one DRAFT or UNPOSTED journal entry, so its amounts count in the ledger balances; other statuses return 409. Posting does not check the accounting period, but a later unpost does. POSTED is Entendre-side only: delivery to the GL is a separate result, shown by is_sync and latest_gl_sync_attempt on the entry. The posted entry.

preview_cash_application

ChatGPT
Preview cash-application allocations (stateless) Returns suggested invoice allocations for a deposit with a confidence score, and writes nothing: exact (one invoice, 1.0), multi_invoice (several invoices, 0.9) or tolerance (one invoice within $50, 0.7). The deposit must be an incoming DEPOSIT, INCOME, REFUND or CHARGEBACK from a bank, Stripe, exchange or wallet source, not yet posted and not claimed; otherwise the 400 reason names the problem. stripe_customer_id narrows the invoices. Candidate allocations with confidence scores.

read_memory

ChatGPT
Read a resolved memory document Reads one memory document (kind=document, the default; path required) or one operational rule (kind=rule; rule_id required, from list_memory; an unknown id returns 404). A document resolves from the organization layer first, then the firm; the response identifies the selected layer. A superseded file with an empty body reads as suppressed: true and stops the fallback, so the firm file below it does not apply. A denied or truncated read does not mean the document is absent. Policy documents do not change live accounting enforcement. The document content and its layer, or one rule.

refresh_cash_application_invoice

ChatGPT
Refresh a Stripe invoice Refresh one invoice directly from Stripe, including manually created invoices. Requires write:cash-application and an active connection with cash application enabled. Send connection_id when several enabled connections exist. Refresh can move an invoice from an inactive or deleted connection after Stripe verifies it. This operation updates invoice data only. Cash application books revenue directly; refresh does not run the AR invoice import or create journals. Previous invoice values have no general undo. The reply is the same invoice row as list_cash_application_invoices. data is the refreshed invoice row, as in list_cash_application_invoices: id (local inv_ ID), status, amount_due, eligible and ineligibility_reason.

refresh_treasury_balances

ChatGPT
Refresh the stored treasury balances Starts a refresh of the stored treasury read that get_treasury_balances serves (wallets and banks; exchanges come from their daily snapshot). Returns request_seq at once. Read get_treasury_balances again until context.refresh.completed_seq (or failed_seq) reaches it. A refresh at MoonPay size takes about a minute. Repeated calls are cheap: one refresh runs per organization, and a call during a running refresh queues one more after it. request_seq: the number of this request. Nothing is read yet when it returns.

remove_account_mapping

ChatGPT
Remove account mapping Removes one account mapping on an active accounting connection: it clears the link (the external id becomes empty) and deletes neither the Entendre account nor the imported GL account. Repeating it changes nothing, and a mapping on an archived account can still be removed. Bank, card and exchange connection ids return 404. Success with no body.

remove_entity_mapping

ChatGPT
Remove entity mapping Removes one entity mapping on an active accounting connection: it clears the link and deletes neither the legal entity nor the imported GL entity. Remove that connection's account mappings first. Repeating it changes nothing. Bank, card and exchange connection ids return 404. Success with no body.

remove_team_member

ChatGPT
Remove team member Removes a member from the firm: it soft-deletes the membership and its client access, and keeps the user account. Needs firm write access and a firm administrator. The last firm administrator cannot be removed; that returns 409. A 409 for unclassified historical client access requires reconciliation; repeating the unchanged request does not resolve it. Success with no body, or 409 for the last administrator or an access-reconciliation conflict.

reopen_accounting_period

ChatGPT
Reopen an accounting period Reopens a closed or soft-closed period so entries can post into it again. It removes the close lock, so the period's balances and later reports can change. It needs elevated permissions. The reopened period.

resend_invitation

ChatGPT
Resend invitation Sends a new link for an existing invitation; the new link replaces the old one and expires after 14 days, and the send cannot be undone. Needs firm write access, a firm administrator. email_sent records whether the email provider accepted the send. An idempotency_key is required for this request. Reuse it only when retrying the same request. The invitation and its delivery result.

resend_request

ChatGPT
Resend request Sends an existing integration request again and keeps the same request and pending items; the send cannot be undone. Needs firm write access. The same key applies only to the same request body. email_sent records whether the email provider accepted the send. An idempotency_key is required for this request. Reuse it only when retrying the same request. The same request and its delivery result.

reverse_accrual

ChatGPT
Reverse a vendor accrual Reverses a posted accrual with a compensating journal entry (debits and credits swapped). The original entry is marked reversed and links to the reversal; nothing is deleted. reversal_date must be today's UTC date, so a past or future date is rejected. reversal_date → the date part of "Now (UTC)" in the session context reason → the user's words; with none given, a short neutral reason ("Reversed at user request") The reversal result with the new journal entry.

reverse_journal_entry

ChatGPT
Reverse a journal entry Reverses one POSTED journal entry by creating a mirror entry; the original becomes REVERSED and stays visible for audit. A reversal cannot be reversed, and a draft cannot be reversed. To detach an entry without new lines, unpost it instead. The original entry and its mirror.

revoke_invitation

ChatGPT
Revoke invitation Revokes a pending invitation: it soft-deletes the invitation. A member who already joined stays; use remove_team_member for that. Needs firm write access and a firm administrator. Success with no body; the invitation is soft-deleted.

search_memory

ChatGPT
Document search is paginated: an empty matches array with next_cursor is incomplete. Continue with that cursor and the same filters. truncated can also indicate omitted matches within files; a null cursor does not remove that limitation. root narrows the search to a known directory, such as knowledge/ for knowledge notes; it does not search other folders. Search organization memory by literal query Searches memory documents (kind=document, the default) or operational rules (kind=rule); query is required. Document search matches a literal full name, not a fuzzy or partial one, and returns matching lines with their path and layer (organization or firm). When a zero-match query has a plural or singular twin, other_number holds the lines the other form matched: those lines are alternative matches, not proof of the requested policy. A match is an excerpt; read_memory returns the document for conflict and supersession checks. Rule search matches a case-insensitive substring across the rule fields. Policy can exist in both documents and rules; no rule does not mean no policy. Matching lines with path and layer, or the complete matching rule list.

set_account_mapping

ChatGPT
Set account mapping Maps one Entendre ledger account to a GL account on an active accounting connection. Copy external_id from list_external_accounts for the same connection; the ledger account must belong to this organization. It changes only the GL sync mapping, not a source's payment-account mapping. Bank, card and exchange connection ids return 404. The saved mapping.

set_entity_mapping

ChatGPT
Set entity mapping Maps one Entendre legal entity to a GL entity on an active accounting connection. Copy external_id from list_external_entities for the same connection; the legal entity must belong to this organization. It creates no entity in the GL. Bank, card and exchange connection ids return 404. The saved entity mapping.

set_organization_purpose_account

ChatGPT
Set realized gain or loss account Updates the ledger account used for realized gains and losses. Only REALIZED_GAIN_LOSS is accepted. Requires write:organizations and a user-owned key. Replaces the current assignment. Existing journal entries stay unchanged. Read the current assignment and verify an active Income or Expense account in the selected organization. Ask the user to approve the organization and target account before replacing the assignment. The saved purpose, ledger_account_id, ledger_account_name and sequence_number inside data.

sync_journals

ChatGPT
Bulk-sync journal entries across periods Synchronizes selected journal entries out to the connected GL (QuickBooks, Xero, NetSuite or DualEntry). It groups them by legal entity and period and queues one sync job per group, so work can still be pending when it returns. The supported workflow selects POSTED or REVERSED entries. Drafts and unknown IDs are not eligible; synchronization does not authorize posting them. Dry-run eligible ids passed routing checks only; they are not a promise that the upload succeeds. Accepted groups and their jobs; GL work can still be pending.

unapply_cash_application

ChatGPT
Reverse an applied cash-application back to draft Returns an applied, not yet Stripe-synced cash application to draft: it unposts the journal entry, removes the deposit's accounting and keeps the allocations for editing. A synced row cannot be unapplied; a refund is needed, which is not available here. The optional reason goes to the cash_application.unapplied webhook. The match in draft with its allocations.

unpost_journal_entry

ChatGPT
Unpost a journal entry Returns a POSTED or REVERSED entry to UNPOSTED and detaches it from its transactions, which become unaccounted and lose their rule match. A POSTED entry also loses its balance impact and asset assignments; a REVERSED entry does not, because its reversal already offset it. If the entry was synced or imported, its GL copy is removed first. Other statuses, or a closed period, return 409. The unposted entry.

update_asset

ChatGPT
Update an asset Updates a manual asset lot; lots from wallet or exchange ingestion cannot be edited. A quantity or cost-basis change adjusts the ledger balance in the same operation. Lot details are available through list_assets. The updated lot and its balance adjustments.

update_client

ChatGPT
Update client Updates a client's fields, or activates a draft client with status: "active" in a separate request. Needs firm write access and a firm administrator. Activation is a separate requested action. A firm API key cannot archive or restore a client (403); the user does that in the dashboard. The updated client.

update_connection

ChatGPT
Update connection Updates a connection's name (1 to 500 characters) or its legal_entity_ids; send at least one, and only one for Stripe or Fireblocks. A legal-entity change also applies to the connection's sources, except for Stripe and Fireblocks, where it applies to new records only. A disconnected bank connection with a saved Plaid item accepts only legal_entity_ids; other disconnected connections return 409. Several NetSuite configurations also return 409; a stale If-Match returns 412. Credentials cannot be changed here. The updated connection.

update_cost_basis_settings

ChatGPT
Update cost basis settings Updates the cost-basis methodology and/or lot-tracking scope for the connected organization. Omitted fields stay unchanged. At least one field is required. This saves settings only; it does not recompute historical gains or losses or create journal entries. Read the current configuration first and explain the requested change. Require the user to choose the method and scope; do not select a tax policy from a demo request. A missing configuration is not permission to configure one. The saved methodology and accounting_based inside data.

update_current_organization

ChatGPT
Update current organization Updates the organization fields you send: name, timezone, time_format, web_address or logo_url. Send at least one field; omitted fields stay the same. A timezone change affects more than the display label. The updated organization.

update_firm_settings

ChatGPT
Update firm settings Updates the firm name or the Entendre subdomain slug; send at least one. Needs firm write access and a firm administrator. Client-restricted keys cannot call it. The updated firm settings.

update_journal_entry

ChatGPT
Update a journal entry Updates one entry. DRAFT and UNPOSTED entries accept memo, accounting_date, tag_ids, legal_entity_id, transaction_id and lines (all lines, balanced). POSTED and REVERSED entries accept only memo, tag_ids and line memos and tags: send every line with its id, and keep its account, amount and side. An entry in a Closed period rejects every change, and the status never changes here. A date in a closed period moves the entry to the current open period and returns period_auto_reassigned: true with a warning. The updated entry and any reassignment warning.

update_ledger_account

ChatGPT
Update a ledger account Updates the fields you send on a ledger account: name, type, sequence, parent, clearing flag or asset type. A parent change rejects cycles and recalculates the postable flags. Changing type changes the normal balance of existing postings. A new account and reclassification are separate alternatives to an in-place type change. The updated account.

update_tag

ChatGPT
Update a tag Updates a tag's label. Only value can change; the key is fixed because transactions, journal entries and reports use it. Archive and unarchive are not available here; the user does them in the app. The updated tag.

update_team_member

ChatGPT
Update team member Updates a firm member's role. Needs firm write access and a firm administrator. The last administrator cannot be demoted. A 409 for unclassified historical client access requires reconciliation; repeating the unchanged request does not resolve it. The updated membership.

update_transaction

ChatGPT
Update a transaction Updates one transaction: memo, transaction_type, or spam. transaction_type labels the economic event only; it does not classify the transaction or create a journal entry (classification and category are old names for it, so send one of the three). spam: true marks junk such as scam tokens, dust or fake airdrops through the spam workflow; send it on its own, without memo or type. The updated transaction.

update_vault_document

ChatGPT
Use this to rename a Vault document or move it to another folder. Update a document Renames a Vault document (filename) or moves it to another folder (file_directory). It does not replace the file. It does not change the extracted bill, receipt or statement data. The updated document.

upload_vault_document

ChatGPT
Upload a document Uploads one document to the Vault as base64. Production refuses a request body over 32 MiB, so send files up to about 24 MB. Accepted files: PDF, PNG, JPEG, CSV (a bank statement or transaction list) and XLSX or XLS (stored only); other types are rejected. The server classifies the file, and a document classified as a bill can also post a journal entry. It needs write:vault and a user-linked API key. Uploading is a separate action from mentioning or attaching a file. The created document, with its id and extracted summary.

void_cash_application

ChatGPT
Void a draft or applied cash-application Voids a cash application so it can no longer be applied. A draft voids with no accounting change; an applied, Stripe-pending row also has its journal entry unposted. A synced row cannot be voided, and voiding is not a refund. Allocations stay for audit. The voided match.

write_memory

ChatGPT
Create or update an authored memory document Creates or replaces a memory document (kind=document, default) or upserts an operational rule (kind=rule). A document write replaces the whole file and needs content, rationale and expected_version: read the path with read_memory and pass its version, or null to create (also for an organization override of a firm file); a mismatch returns 409. A policy/ document must open with frontmatter carrying id, scope, authority: binding and status; an empty body with status: superseded suppresses the file. Writable: profile.md, knowledge/ (a new note: knowledge/<slug>.md), policy/, entities/, sources/, ledger/, counterparties/, people/ and skills/<slug>/SKILL.md. Not writable: core.md, notes.md, JSON, knowledge/facts/, knowledge/preferences/, knowledge/decisions/, timeline/, workstreams/, refs/, journal/, artifacts/. A rule write omits path: classification needs category_ledger_account_id and one of merchant_name, source_category, blockchain_address or description_contains (literal text); cash application needs title, statement and matcher, and route_to_account also needs ledger_account_id and ledger_account_name. A saved rule applies to later runs only; to classify rows now, use classify_transactions. The new document version, or a rule outcome of created, already_exists or replaced.

execute_code

Claude

list_directory

Claude

read_file

Claude

Similar Apps

Listed on MCP App Store

Listed on MCP App Store

Add this badge to your site or README. It links to this page.

HTML

<a href="https://mcp-app-store.keenethics-labs.com/app/entendre"><img src="https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg" alt="Listed on MCP App Store" width="220" height="44"></a>

Markdown

[![Listed on MCP App Store](https://mcp-app-store.keenethics-labs.com/badges/listed-on-mcp-app-store.svg)](https://mcp-app-store.keenethics-labs.com/app/entendre)