activities_create
ChatGPTLogs an activity on a deal, a contact, or both. activity_type (note, call, email, meeting or task) and title are required. description, activity_date, deal_id and contact_id are optional. Put the substance in description rather than title: it is returned in full by activities_list and is what a later conversation will read back, so a well written note is what makes the CRM useful as memory. activity_date takes an ISO 8601 timestamp with offset, e.g. 2026-11-15T10:00:00+02:00; leave it out to default to now, which is usually correct. Creates a new entry every time. Returns the created activity.
activities_list
ChatGPTLists timeline activities, newest first. Filter by deal_id, contact_id or activity_type, or fetch one by id. Descriptions are returned in full, so this is the tool for reading what a lead actually said on a call or in a meeting. Results also include status_change entries that Fluid writes automatically when a deal is won, lost or moved between stages; filter activity_type = status_change to see only those (they cannot be created or edited). Returns { items[], next_cursor, has_more }.
activities_update
ChatGPTEdits one activity by id. Partial: only the fields you pass change. activity_date takes an ISO 8601 timestamp with offset. Safe to retry. Returns the updated activity.
companies_bulk_create
ChatGPTImports up to 50 companies in one call using Fluid's duplicate matching: a row whose website host or name already exists is reported as matched and not created, and www is stripped before matching, so www.example.com finds example.com. A row with no name is skipped and listed in row_reports (reason missing_name); the valid rows still import. Retrying the same rows is always safe. name is required per row. A row may carry label ('Partner|Key Account' attaches both; a name not in the catalog is added to it). For more than 50 rows, split into chunks and call once per chunk. Records created here cannot be deleted through this API: confirm with the user before importing a large list. Returns counts (received, created, skipped_invalid, matched_existing, duplicates_in_batch, failed), created_companies[{row_idx,id}] and row_reports[{row_idx,reason}], numbered by your rows counting from 1 (row_idx 1 is your first row).
companies_create
ChatGPTCreates one company. name is required. This does not check for duplicates: calling it twice with the same company creates two records. Prefer companies_bulk_create even for a single company, because that one matches on website and name first. website is stored exactly as given and is used to match duplicates later, so a malformed website becomes a matching key. label attaches labels ('Partner|Key Account' attaches both; a new name is added to the catalog). Records created here cannot be deleted through this API: confirm with the user before creating records they did not explicitly ask for. Returns the created company with labels and custom_fields.
companies_list
ChatGPTFinds companies. Filter by text (q matches name, website and industry) or updated_since (an ISO 8601 timestamp), or fetch specific companies by id or ids. Every row carries its labels (names) and custom_fields. limit is capped at 500 and the applied page size is echoed as limit. Returns { items[], next_cursor, has_more, limit }; pass cursor to get the next page and stop when has_more is false (an unreadable or altered cursor is rejected with an error rather than silently restarting from page 1: drop it and start again).
companies_update
ChatGPTUpdates one company by id. Partial: only the fields you pass change. custom_fields sets values by field name, and null clears a value. label replaces the company's whole label set ('Partner|Key Account' attaches both, a new name is added to the catalog, empty string or null removes all); omit it to leave labels untouched. Safe to retry. Returns the updated company.
contacts_bulk_create
ChatGPTImports up to 50 contacts in one call using Fluid's duplicate matching: a row whose email, or name within its company, already exists is reported as matched and not created. A company named in a row is matched or created and linked automatically. A row with no name is skipped and listed in row_reports (reason missing_name); the valid rows still import. Retrying the same rows is always safe. name is required per row. For more than 50 rows, split into chunks and call once per chunk. Records created here cannot be deleted through this API: confirm with the user before importing a large list. Returns counts (received, created, skipped_invalid, matched_existing, duplicates_in_batch, failed, companies_created, companies_matched), created_contacts[{id,name,email}] and row_reports[{row_idx,reason}], numbered by your rows counting from 1 (row_idx 1 is your first row).
contacts_create
ChatGPTCreates one contact. name is required. This does not check for duplicates: calling it twice with the same person creates two records. Prefer contacts_bulk_create even for a single contact, because that one matches on email first. Text fields are trimmed of surrounding spaces; email keeps its case and is used to match duplicates later. expected_close_date takes a calendar date as YYYY-MM-DD. Records created here cannot be deleted through this API: confirm with the user before creating records they did not explicitly ask for. Returns the created contact with labels and custom_fields.
contacts_list
ChatGPTFinds contacts. Filter by text (q matches name, email and company), organization_id or updated_since (an ISO 8601 timestamp), or fetch specific contacts by id or ids. Every row carries its labels (names) and custom_fields. limit is capped at 500 and the applied page size is echoed as limit. Returns { items[], next_cursor, has_more, limit }; pass cursor to get the next page and stop when has_more is false (an unreadable or altered cursor is rejected with an error rather than silently restarting from page 1: drop it and start again).
contacts_update
ChatGPTUpdates one contact by id. Partial: only the fields you pass change. label attaches or replaces a label, custom_fields sets values by field name (null clears a value), organization_id links a company. expected_close_date takes YYYY-MM-DD. Safe to retry. Returns the updated contact.
deals_bulk_create
ChatGPTImports up to 50 deals in one call with Fluid's duplicate matching: companies by website then name, contacts by email or by name within the company, duplicate rows in the batch collapsed, and a row whose contact already has a deal of that name reported as matched, not created. A row with no contact_name is skipped and listed in row_reports (missing_name); the rest still import. pipeline_id is required; each row needs contact_name and deal_name. A row may carry label ('Hot|Warm' attaches both; a new name joins the catalog), reminder_date (YYYY-MM-DD, creates a reminder at 09:00 UTC that day), expected_close_date (YYYY-MM-DD) and custom_fields by name; a custom_fields value that does not fit its field type is left blank and reported (cf_type_mismatch with field and expected), the row still imports. Retrying the same rows is safe with on_existing skip, the default. Over 50 rows: split into chunks. Records created here cannot be deleted through this API: confirm before importing a large list. Returns counts (received, created, skipped_existing, skipped_invalid, duplicates_in_batch, contacts_new, contacts_reused, reminders_created), created_deals[{row_idx,id}], matched_existing[] and row_reports[{row_idx,reason}] numbered by your rows counting from 1 (row_idx 1 is your first row); reason is matched_existing, existing_deal_found (created again under on_existing create), within_batch_duplicate, missing_name or cf_type_mismatch.
deals_create
ChatGPTCreates one deal. title and stage_id are required; get stage ids from workspace_schema. This does not check for duplicates; to add many leads with duplicate protection use deals_bulk_create. expected_close_date takes a calendar date as YYYY-MM-DD. value must be below 10 billion, and setting it also writes the same number onto the linked contact's value field, so mention that if the user tracks contact values separately. Records created here cannot be deleted through this API: confirm with the user before creating records they did not explicitly ask for. Returns the created deal with its labels and custom_fields.
deals_list
ChatGPTFinds deals. Filter by text (q matches title and deal name), stage_id, contact_id, organization_id, status (active, won or lost) or updated_since (an ISO 8601 timestamp), or fetch specific deals by id or ids. Every row carries its labels (names) and custom_fields. limit is capped at 500 and the applied page size is echoed as limit. Returns { items[], next_cursor, has_more, limit }; pass cursor to get the next page and stop when has_more is false (an unreadable or altered cursor is rejected with an error rather than silently restarting from page 1: drop it and start again).
deals_update
ChatGPTUpdates one deal by id. Partial: only the fields you pass change. stage_id moves it, status wins, loses or reopens it, label attaches or replaces its label, custom_fields sets values by field name (null clears). Closing (status won or lost) permanently deletes every reminder on the deal, done or not; reopening does not bring them back. That is Fluid's design, not an error, and the app behaves the same way. The result includes deleted_reminders, the number of reminders that were cleared, so the count can be reported to the user. expected_close_date takes YYYY-MM-DD. value must be 0 or more and below 10 billion, and also writes the same number onto the linked contact. A deal's pipeline is its stage's pipeline, so a stage_id from another pipeline moves the deal there; custom field values of the old pipeline are kept, hidden until it moves back. Unlike the app, this does not check the new pipeline's required fields: check workspace_schema for fields marked required and fill them first. A workspace automation may forward a deal from a trigger stage, so read stage_id back from the result. Safe to retry. Returns the updated deal, plus deleted_reminders when this update closed it.
get_pipeline_summary
ChatGPTAnswers 'how many deals are in X' and 'what is the pipeline worth' without paging through deals. Returns active deal count and total value per stage per pipeline, won and lost totals, workspace-wide deal totals, reminder counts (overdue, due today, due this week) and contact and company counts. Always pass tz as the user's own timezone in IANA form, e.g. Europe/Helsinki; it defaults to UTC, which changes what counts as due today and due this week.
reminders_create
ChatGPTCreates a reminder on a deal. deal_id and reminder_at are required, note is optional. reminder_at takes an absolute ISO 8601 timestamp including the user's UTC offset, e.g. 2026-11-15T10:00:00+02:00. Work out anything relative the user says, such as 'next Friday' or 'in two weeks', into an absolute timestamp yourself before calling. Prefer a midday time over midnight. If you only have a date, do not invent a time: send it as YYYY-MM-DD together with tz (the user's IANA timezone) and it is stored at 12:00 that day in that zone, so it never lands on the wrong day. If you do not know the user's timezone, ask for it rather than assuming UTC. Returns the created reminder.
reminders_list
ChatGPTReturns the user's open reminders with the deal name and the deal's contact name and email already attached. Use it for 'what is due today', 'what is overdue' and 'what is due this week'. Always pass tz as the user's own timezone in IANA form; it defaults to UTC, which shifts every bucket. due filters by the reminder's date in that timezone: today, overdue, week (today through Sunday), future (after today), or all. Sorted by reminder_at ascending. Each returned row carries its exact bucket in due (today, overdue, week = later this week, future = after this week), and the filter of the same name always includes that row, so a value read from a row can be passed straight back. There is no cursor: raise limit (max 500) if truncated is true.
reminders_update
ChatGPTUpdates one reminder by id: a new reminder_at snoozes it, and note and deal_id can also change. There is no 'done' state for reminders in Fluid; when the user has handled a follow-up, either snooze it or leave it, and tell them reminders are cleared in the app. reminder_at takes an absolute ISO 8601 timestamp with the user's offset; a bare YYYY-MM-DD plus tz is stored at local midday. A reminder on a deal that has since been marked won or lost no longer exists and returns not found. That is expected, not a temporary failure: do not retry it. Safe to retry otherwise. Returns the updated reminder.
workspace_schema
ChatGPTDescribes a workspace in one call: pipelines with their stages (ids, order, win probability), the label catalog, and the custom field definitions for deals, contacts and companies. Call it before creating or moving deals (you need a stage_id), attaching labels, or writing custom_fields (matched by field name). Two things to know before writing. Deal custom fields listed with a pipeline_id belong to that pipeline: set them only on deals in it (the app shows the value only while the deal is in that pipeline). Fields with required true must be filled before a deal is moved into their pipeline; deals_update does not enforce this, the app does. Label names are matched without case sensitivity, and a name that does not exist yet is created as a new label that cannot be deleted through this API, so check the catalog before inventing one. Returns { pipelines[{id,name,stages[]}], labels[], custom_fields{deal[],contact[],company[]} }.
workspaces_list
ChatGPTLists the Fluid workspaces the signed-in user belongs to. Use it first when a call fails asking for workspace_id, or when the user mentions a specific workspace. Returns id, name and the user's role for each. When there is exactly one, it is the default for every other tool.
activities_create
Claudeactivities_list
Claudeactivities_update
Claudecompanies_bulk_create
Claudecompanies_create
Claudecompanies_list
Claudecompanies_update
Claudecontacts_bulk_create
Claudecontacts_create
Claudecontacts_list
Claudecontacts_update
Claudedeals_bulk_create
Claudedeals_create
Claudedeals_list
Claudedeals_update
Claudeget_pipeline_summary
Claudereminders_create
Claudereminders_list
Claudereminders_update
Claudeworkspace_schema
Claudeworkspaces_list
Claude