Affinity CRM MCP server
Yes. Affinity CRM publishes an official MCP server at https://mcp.affinity.co/mcp. It's hosted and needs an OAuth app that you set up in Affinity CRM first. It offers 97 tools. We last checked it on 22 Nov 2026.
Official: answers MCP requests at mcp.affinity.co, on Affinity CRM's own domain affinity.co. Checked 22 Nov 2026. Maintained by Stuart Brameld.
Key facts
As of 22 Nov 2026, Affinity CRM's official MCP server scores 75/100 on MartechDB's Agent Readiness Index (v1.1), joint 27th of 37 in CRM. It answered 56 of our 56 checks in the last 90 days and lists 97 tools.
How the Agent Readiness Index works · Wrong? Suggest a correction
Affinity CRM MCP server review
75/100
Affinity CRM's MCP server scores 75/100 on MartechDB's Agent Readiness Index: easy to connect an AI agent to, with a few gaps. That ranks it joint 27th of 37 in CRM.
The good
Hosted by Affinity CRM: connect by URL, with nothing to install
Documented, with 97 tools known
Answered 100% of our checks in the last 90 days
The bad
Signing in needs an app registered with Affinity CRM first
Needs a paid plan, with no free trial
Agent access costs more than the base plan
Official MCP server
Affinity CRM MCP server
- Address
- https://mcp.affinity.co/mcp
- Runs
- Hosted
- Sign-in
- Cost to use
- Plan: Scale, Advanced, and Enterprise plans · No free access · Costs more than the base planFrom Affinity CRM's own page, pricing from this page
- Official MCP Registry
- Not listed
- How we know it's official
-
- Answers MCP requests at mcp.affinity.co, on Affinity CRM's own domain affinity.co.
- Documentation
- developer.affinity.co/pages/mcp/introduction
- Tools
- 97 tools · listed 22 Nov 2026
- First seen
- 1 Oct 2026
- Last checked
-
Reachable ·
16 Nov 2026: reachable17 Nov 2026: reachable18 Nov 2026: reachable19 Nov 2026: reachable20 Nov 2026: reachable21 Nov 2026: reachable22 Nov 2026: reachable
- Uptime
- 100% · 56 of 56 checks reachable since 1 Oct 2026
Affinity CRM MCP server tools (97)
As the server lists them. Names and descriptions are the vendor's own.
-
get_current_userGet information about the current user. Use this tool to verify authentication and understand available API access levels. Returns: Information about the user, their current organization, and API key permissions, plus `access`: "read-write" or "read-only". Report `access` as the user's access level. Do not infer it from individual scope names: "mcp.read" is a sub-scope that is also granted alongside the read-write "mcp" scope. `access` is absent when the grant carries no recognized scope.
-
search_companies_top_matchesSearch companies using natural language, structured criteria, or both simultaneously. When query and search_criteria are provided, both semantic search and structured search run in parallel and results are merged. Call get_company_info for richer field data on results. DEFAULT for all company lookups, including structured-only filters with no semantic query. Only route to search_all_companies() when the request contains an explicit exhaustive-intent word such as "all", "every", "full list", "complete list", or "export", or when the user needs field data returned inline with results or a total count. Use this tool when discovery phrasing is present ("what companies", "which companies", "find companies", "show me companies"). RULE 1 - EXACT NAME OR DOMAIN LOOKUP (check this FIRST): If the request is to find one specific company by its name or domain ("find Stripe", "look up stripe.com", or resolving a company name to an ID before calling another tool), pass ONLY search_criteria and OMIT…
-
search_all_companiesPaginate exhaustively through companies using structured filters, sorts, and/or a keyword. Use this tool when: - The request signals exhaustive intent ("all", "every", "full list", "export") - The user needs field data returned inline with results (field_ids or field_types) - The user needs a total count of matching companies (total_count=True) - Paginating through a complete result set from a previous call For all other searches, including structured-only filters, use search_companies_top_matches() instead. Omitting search_criteria returns every company with default pagination. FIELD SELECTION: field_ids and field_types are mutually exclusive; passing both raises ValueError. Valid field_types are enriched, global, and relationship-intelligence. When both are omitted, only company identity (id, name, domain, domains, isGlobal) is returned. DISCOVERING FIELDS: Use get_entity_fields(entity_type=EntityType.COMPANY, includes=[FieldInclude.FILTERABILITY, FieldInclude.SORTABILITY]) to find…
-
get_company_infoReturns basic information and non-list specific field data on the requested company. By default, returns all field data including location, description, and more. Use get_entity_fields tool if you need to filter to specific fields. IMPORTANT: field_ids and field_types are mutually exclusive. They cannot be used together. Args: company_id (required): The ID of the company field_ids: List of field IDs to return. field_types: List of field types to return. Defaults to ['enriched', 'global', 'relationship-intelligence'] Returns: Metadata about the specified company including field data. A field on an opportunity you can't manage may come back with type "hidden" and a masked value.
-
get_company_list_entriesFetches list entries for the given company. Args: company_id (required): The ID of the company cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, max is 100 Returns: A dict with `data` (list of list entries) and `pagination` with `nextCursor` / `prevCursor` strings (pass directly to the `cursor` param; absent when there is no such page) and `nextUrl` / `prevUrl` full URLs (null when there is no such page).
-
get_merge_taskChecks whether a company or person merge has finished, by its task id. Use after merge_companies or merge_persons returned with confirm=true: both return a task_id, and this reads GET /v2/tasks/<entity>-merges/{task_id} once. It does not poll. A merge can stay "in-progress" for several minutes while Affinity retries it; read again when the user asks. "failed" is final. IMPORTANT: entity_type must match the tool that produced the task: a company task read as "person" (or the reverse) is a 404, the same as an unknown id. Requires the Manage Duplicates permission and organization admin role, like the merge tools; a token without it gets a 403. Args: task_id (required): The task_id returned by merge_companies or merge_persons, a UUID such as "76251d92-99e1-416f-a3d2-0935435eb672". Anything that is not a UUID raises ValueError before any API call. entity_type (required): "company" for a merge_companies task, "person" for a merge_persons task. Returns: {"task_id", "entity_type", "status", "…
-
render_market_mapRender an investment thesis market map as an interactive scatter plot and coverage tables. IMPORTANT: only call this after searching Affinity (and optionally the web) for companies matching the thesis and scoring each for thesis fit, per the market-map-generator skill. This tool does not search, score, or classify anything itself. It draws exactly the companies it is given, so pass the same de-duplicated, quality-filtered results already shown to the user in text. Never invent a company, score, or funding figure not already in those results. Args: thesis: The investment thesis being visualized. your_companies: Companies with interaction history ("tracked" or "org-relevant" source only), plotted on the scatter and listed in the "Your Companies" table. Maximum 20. explore_companies: Companies with no interaction history ("global-cold" or "web" source only), listed in the "Companies to Explore" table only. Maximum 20. Must not repeat a domain already present in `your_companies` - de-dupl…
-
get_entity_fieldsReturns metadata on non-list specific company or person fields. Args: entity_type: Type of entity to return fields for. Options: 0 = person, 1 = company cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, max is 100. filter: Filter fields by name. `name` is the only filterable property. Two operators, both case-sensitive, so match the field's actual capitalization: - Exact match: name="Location" - Substring match: name=~"Funding" Wrap any value containing spaces in double quotes, e.g. name=~"Last Funding" Combine clauses with & (AND), | (OR), ! (NOT) and parentheses, e.g. name=~"Funding" | name="Location" includes: Extra per-field metadata to return, omitted by default. Request these before building a search_companies_top_matches, search_all_companies, or search_persons filter or sort to confi…
-
get_list_fieldsReturns metadata on fields available for a given list. Args: list_id: ID of list to return fields for. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, max is 100. filter: Filter fields by name. `name` is the only filterable property. Two operators, both case-sensitive, so match the field's actual capitalization: - Exact match: name="Location" - Substring match: name=~"Funding" Wrap any value containing spaces in double quotes, e.g. name=~"Last Funding" Combine clauses with & (AND), | (OR), ! (NOT) and parentheses, e.g. name=~"Funding" | name="Location" includes: Extra per-field metadata to return, omitted by default. Request these before building a search_list_entries filter or sort to confirm the field supports the operator and ordering you intend. - FieldInclude.FILTERABILITY: how the fi…
-
get_list_field_dropdown_optionsReturns the dropdown options for a specific dropdown, ranked-dropdown, or status-dropdown field on a list. Use the returned dropdown option IDs when writing dropdown field values via upsert_list_entry_field_values. A status-dropdown option additionally carries `rank`, `color`, `statusCategory` (one of "open", "won", "lost", "on-hold"), and `winRate` (0-100, only set when statusCategory is "open"). Args: list_id: ID of list field_id: human-readable string ID of field to return options for (ex. "field-123") cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 20, maximum is 100. Returns: A dict with `data` (metadata about each dropdown option) and `pagination` with `nextCursor` / `prevCursor` strings (pass directly to the `cursor` param; absent when there is no such page) and `nextUrl` / `prevUrl` ful…
-
get_entity_field_dropdown_optionsReturns the dropdown options for a dropdown, ranked-dropdown, or status-dropdown field on a company or person. Use the returned dropdown option IDs when writing dropdown field values via upsert_entity_field_values. Company and person fields are global to the org, so options are resolved from the field alone; no entity ID is needed. A status-dropdown option additionally carries `rank`, `color`, `statusCategory` (one of "open", "won", "lost", "on-hold"), and `winRate` (0-100, only set when statusCategory is "open"). Args: entity_type: 0 = person, 1 = company. field_id: human-readable string ID of field to return options for (ex. "field-123") cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 20, maximum is 100. Returns: A dict with `data` (metadata about each dropdown option) and `pagination` with `…
-
get_field_value_changesGet a paginated list of field value changes across the organization. Designed for delta sync: bound the first call with a `changedAt` filter, then store `pagination.nextCursor` and pass it to subsequent calls to retrieve only new changes since the last sync. The scope is the whole organization and a change's `value` can be a text field, so an unfiltered first call sweeps the org's entire edit history at 100 rows a page. Narrow by `changedAt` rather than paging to the start of time, and do not sweep unprompted — report what the first page holds and offer to fetch more. To get the change history of one company, person, or opportunity: 1. Find the record. If the user gives the record type, use only the search tool for that type: `search_companies_top_matches`, `search_persons`, or `search_opportunities`. If the user gives only a name, search all three types. If more than one record matches, stop and ask the user which record they mean before step 2. Ask with the AskUserQuestion tool if y…
-
get_suggested_field_updatesGet one set of suggested field updates (SFUs) with all its suggestions. A set is one Data Update Agent run over a note or meeting. Each suggestion proposes a new value for one field on a company, person or opportunity. Use list_suggested_field_updates to find set ids. If the note or meeting id is known, pass subject_type and subject_id to list only that subject's sets. IMPORTANT: currentValue is the field's value when the suggestions were generated. The record may have changed since. Args: suggested_field_update_id (required): The set id, a UUID such as "7c9e6679-7425-40de-944b-e07fc1f90ae7". Returns: The set. status "pending" is still generating and has no suggestions; "open" waits on the user. A run that found nothing is "done" with no suggestions. Show records as "Acme Corp (company 4521)". Errors: 404 when the set does not exist, is not this user's, or the user has no access to suggested field updates.
-
list_suggested_field_updatesList the user's sets of suggested field updates (SFUs), newest first. A set is one Data Update Agent run over a note or meeting. Items have counts but no suggestions: call get_suggested_field_updates with an id for those. IMPORTANT: Runs that found nothing are not listed. An empty list does not mean the agent is off; say there are no suggestions, and that the agent may not be enabled for this user. IMPORTANT: Sets the user cannot see are removed after the page is read, so a page can be short or empty while `pagination.nextCursor` is set. Keep paging until `nextCursor` is null. Args: status: Only sets in this status. "open" sets wait on the user. "pending" sets are still generating. subject_type: Only sets generated from a "note" or a "meeting". subject_id: Only sets for this note or meeting id. Requires subject_type, because a note and a meeting can share an id; without it, raises ValueError before any API call. A note attached to a meeting has its sets under the meeting id. created_a…
-
get_suggested_field_updates_taskCheck whether suggested field update (SFU) generation has finished. Reads the task once. It does not poll. Generation can take several minutes; read again when the user asks. Args: task_id (required): The task id, a UUID such as "7c9e6679-7425-40de-944b-e07fc1f90ae7". It equals the set id. Returns: The task. On "success", call get_suggested_field_updates with suggestedFieldUpdate.id. "failed" is final and carries no reason. Errors: 404 when the task does not exist, is not this user's, or the user has no access to suggested field updates.
-
list_suggested_field_updates_tasksList the user's suggested field update (SFU) generation tasks, newest first. Args: status: Only tasks in this status. limit: Tasks per page, 1 to 100. Default 20. cursor: Pass `pagination.nextCursor` or `pagination.prevCursor` from a prior response, unchanged. Returns: One page of tasks. Errors: 400 for a bad cursor; 404 when the user has no access to suggested field updates.
-
get_investor_executive_connectionsFind warm intro paths to a company through investor-executive connections: investors at your firm who are connected to executives at the target company via a shared portfolio company. These are INFERRED second-degree connections derived from investment relationships, not from your org's own interaction history. The people surfaced may never have emailed or met anyone at your org. For relationship strengths based on your org's actual interactions ("who do we know at this company"), use `get_company_relationships` or `get_person_relationships` instead. IMPORTANT: `company_id` is an Affinity company ID, not a name or domain. Call `search_companies_top_matches` first to resolve a company name to its ID. An empty `data` list means there are no inferred connections for the company. Args: company_id: The ID of the target company whose executives to find connections to. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a pr…
-
get_coworker_connectionsFind warm intro paths to a company through former coworkers: people at your org whose work history overlaps with people currently at the target company. These are INFERRED second-degree connections derived from shared employment history, not from your org's own interaction history. The people surfaced may never have emailed or met anyone at your org. For relationship strengths based on your org's actual interactions ("who do we know at this company"), use `get_company_relationships` or `get_person_relationships` instead. IMPORTANT: `company_id` is an Affinity company ID, not a name or domain. Call `search_companies_top_matches` first to resolve a company name to its ID. An empty `data` list means there are no inferred connections for the company. Args: company_id: The ID of the target company whose people to find former coworkers of. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or…
-
get_intro_pathsFind ranked warm intro paths to a person or company: who at your org can introduce you, directly or through one bridge person, strongest path first. Each path has one or two edges. One edge is a direct relationship between a person at your org and the target. Two edges go through a bridge person, where `edges[0].to` and `edges[1].from` are the same person. Paths combine your org's interaction history and inferred connections (shared work history, investor-portfolio relationships). IMPORTANT: Pass exactly one of `person_id` or `company_id`. Both or neither raises ValueError before any API call. These are Affinity IDs, not names: call `search_persons` or `search_companies_top_matches` first to resolve a name to its ID. IMPORTANT: Paths are already ranked by the API. Present them in the order returned, using `rank` and `tier`. Do not re-rank them. The response carries no absolute path score, and `interactionScore` on an edge is not a path score. When the paths cannot be loaded for the ta…
-
render_intro_paths_graphRender ranked warm intro paths from get_intro_paths as an interactive network-map graph, with an evidence panel for each relationship. Call this every time you present get_intro_paths results to the user: first give the paths as text, then call this tool for the same paths. Do not wait for the user to ask for a map. Also call it when the user asks to see, map, visualize, or diagram the paths. Do not call it on your own when get_intro_paths returned no paths, or returned `unavailable`. Call this only after fetching paths with get_intro_paths. This tool does not fetch or rank anything itself. It draws exactly the paths it is given, so pass the same paths already shown to the user in text, in the same order. Never invent a person, relationship, score, or signal that get_intro_paths did not return. Each relationship is drawn by what supports it: one with a `score_pct` as confirmed and scored (solid line with the number), one with only `linkedin_connected_on` as a confirmed LinkedIn connec…
-
get_listsGet a page of lists in the user's organization that they have access to, optionally filtered by name. More lists may exist than one page holds: keep calling with `pagination.nextCursor` until it is absent to see every list. Args: term: Case-insensitive substring filter on list name. Returns lists whose name contains this string anywhere. Not fuzzy - typos will miss; if results are empty, try a shorter or different keyword. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, max is 100. Returns: A dict with `data` (list metadata such as list name, entity type, creator ID, owner ID, and `list_url`) and `pagination` with `nextCursor` / `prevCursor` strings (pass directly to the `cursor` param; absent when there is no such page) and `nextUrl` / `prevUrl` full URLs (null when there is no such page).
-
get_single_list_entryRetrieve a single list entry from the given list. The list entry contains basic information about the related person/company/opportunity and list-specific field data. By default, returns only list-specific fields. IMPORTANT: field_ids and field_types are mutually exclusive. Passing both raises ValueError. WARNING: List entry may contain large field data. Best practices to minimize response size: - Select specific fields using field_ids or field_types (use get_list_fields tool to identify fields) - Request additional field types (enriched, global, relationship-intelligence) only when needed Args: list_id (required): ID of list list_entry_id (required): ID of list entry field_ids: List of field IDs to return. field_types: List of field types to return. Defaults to ['list']. Available types: 'enriched', 'global', 'relationship-intelligence', 'list' Returns: JSON object for the single list entry: entity info (person/company/opportunity) and the selected list-specific field values. A field…
-
get_list_infoGet metadata on a single list Args: list_id (required): ID of list Returns: JSON object containing list metadata such as list name, entity type, creator ID, and owner ID.
-
search_list_entriesSearch list entries on a list using filters, sorts, and/or a keyword. Omitting `search_criteria` returns every entry on the list with default pagination. WARNING: Lists can contain many entries with large field data. To minimize response size, start with the default limit=10 (increase only if needed), use pagination (cursor) to iterate, and return only the fields you need (see FIELD SELECTION). FIELD SELECTION (which fields to return per entry): `field_ids` and `field_types` are mutually exclusive; passing both raises ValueError. When both are omitted, defaults to `field_types=['list']` (list-specific fields only). Request additional types (enriched, global, relationship-intelligence) only when needed. DISCOVERING FIELDS: Use get_list_fields to find the field IDs you can reference in filters, sorts, and search.fieldIds. Call it with includes=[FieldInclude.FILTERABILITY, FieldInclude.SORTABILITY] to learn which operators each field can be filtered with and whether it can be sorted (inc…
-
get_meetingsGet information about past and future meeting interactions and their attendees. Args: filter_expr: Filter expression using Affinity query language. Available fields: - id: unique identifier for the meeting (operators: =) Example: "id=1" or "(id=1 | id=2)" - startTime: When the meeting was scheduled (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "startTime>2025-01-01T01:00:00Z" - createdAt: When the meeting was created in Affinity (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "createdAt>=2025-01-01T00:00:00Z" - updatedAt: When the meeting was last updated (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "updatedAt<=2025-12-31T23:59:59Z" Complex examples: - Date range: "startTime>=2025-01-01T00:00:00Z & startTime<2025-02-01T00:00:00Z" - Multiple date filters: "startTime>2025-01-01T00:00:00Z & createdAt>=2024-12-01T00:00:00Z" Boolean logic: & (AND), | (OR), () for grouping Full filtering spec: https://developer.affinity.co/pages/external-api-v2…
-
get_meetings_for_entityGet information about meetings for a specific company, person, or opportunity. IMPORTANT: The date range between start_time and end_time cannot exceed 365 days. Be mindful of leap years. A larger range is automatically clamped to the most recent 365 days (start_time moves forward) and the response carries a `warnings` entry stating the window actually used; call again with an earlier end_time to fetch older meetings. start_time must be strictly before end_time: an inverted or empty range is rejected with an error, not clamped. Args: entity_id: The ID of the company, external person, or opportunity. A person must be an external contact or a collaborator: the API rejects an internal person with "Person is not an external or collaborator". entity_type: What type of entity this ID refers to. Options: 0 = person, 1 = company, 8 = opportunity start_time: ISO 8601 formatted timestamp to filter meetings starting from this time. Must be before end_time Format: "2025-01-01T00:00:00Z" end_time:…
-
get_callsGet the logged calls that the current user can see, with their dates and attendees. Each call has a `startTime`: the date and time of the call that the user gave when they logged it. `createdAt` is when the record was created in Affinity, not when the call took place. To select calls by the day they took place, filter on `startTime`. The results include only the calls that the current user has permission to see, so a count of them is not an organization total. Calls come in id order, which is the order in which Affinity stored them, not the order of their dates. The first page holds the calls that Affinity stored first. To get recent calls, filter on `startTime`. To get the calls of one company, person, or opportunity, use get_interactions_for_entity. It returns calls newest first. Args: filter_expr: Filter expression using Affinity query language. Available fields: - id: unique identifier for the call (operators: =) Example: "id=1" or "(id=1 | id=2)" - startTime: When the call took p…
-
get_chat_messagesGet the logged chat messages that the current user can see, with their dates, direction, and participants. The API returns no message text. A user logs a chat message by hand, and the text they enter is kept as a note on the message. get_interactions_for_entity returns the ids of these notes. To read the text of a note, use query_notes with filter_expr "id=<note id>". Chat messages come newest first by `createdAt`, which is when a user logged the message in Affinity, not when it was sent. To select messages by the date they were sent, filter on `sentAt`. To get the chat messages of one company, person, or opportunity, use get_interactions_for_entity. It returns chat messages newest first by the date they were sent. Args: filter_expr: Filter expression using Affinity query language. Available fields: - id: unique identifier for the chat message (operators: =) Example: "id=1" or "(id=1 | id=2)" - sentAt: When the message was sent (operators: >, <, >=, <=) Format: ISO 8601 timestamp Exam…
-
get_emailsGet the emails that the current user can see, with their dates, subjects, direction, and participants. IMPORTANT: The API never returns email bodies, only metadata. If the user cannot see the subject of an email, `subject` is "********". Emails come in id order, which is the order in which Affinity stored them, not the order of their dates. The first page holds the emails that Affinity stored first. To get recent emails, filter on `sentAt`. To get the emails of one company, person, or opportunity, use get_interactions_for_entity. It returns emails newest first. Args: filter_expr: Filter expression using Affinity query language. Available fields: - id: unique identifier for the email (operators: =) Example: "id=1" or "(id=1 | id=2)" - sentAt: When the email was sent (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "sentAt>=2025-01-01T00:00:00Z" - createdAt: When the email was added to Affinity (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "createdAt>=2025-0…
-
get_interactions_for_entityGet the calls, chat messages, or emails (not meetings) of a specific company, person, or opportunity. For meetings, use get_meetings_for_entity. The results include only the interactions that the current user has permission to see. IMPORTANT: The date range between start_time and end_time cannot exceed 365 days. A larger range is clamped to the most recent 365 days (start_time moves forward), and a `warnings` entry in the response states the window that was used. To fetch older interactions, call again with an earlier end_time. start_time must be strictly before end_time: an inverted or empty range causes an error, not a clamp. Args: interaction_type: The kind of interaction to get: - InteractionType.CALL (1) - InteractionType.CHAT_MESSAGE (2) - InteractionType.EMAIL (3) If unclear which type the user means, ask before calling. entity_id: The ID of the company, external person, or opportunity. A person must be an external contact or a collaborator: the API rejects an internal person w…
-
query_notesFind notes in Affinity. Filter by ID, creator, and/or creation date. Args: filter_expr: Filter expression using Affinity query language. Available fields: - id: unique identifier for the note (operators: =) Example: "id=123" or "(id=1 | id=2 | id=10)" - creator.id: Person ID that created the note (operators: =) Example: "creator.id=456" - createdAt: When the note was created in Affinity (operators: >, <, >=, <=) Format: ISO 8601 timestamp Example: "createdAt>=2026-01-01T00:00:00Z" Complex examples: - Date range: "createdAt>=2026-01-01T00:00:00Z & createdAt<2026-02-01T00:00:00Z" - By creator in date range: "creator.id=123 & createdAt>=2026-01-01T00:00:00Z" Boolean logic: & (AND), | (OR), () for grouping Full filtering spec: https://developer.affinity.co/pages/external-api-v2/filtering cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: Number of items t…
-
get_notes_for_entityGet a page of notes attached to a specific company, person, or opportunity. For the reverse (finding entities attached to a note), use get_entities_attached_to_note. The entity may have more notes than one page holds. For counts and totals, pass total_count=True rather than fetching rows: note bodies are free text and 20 of them have measured over 100k characters, past the tool-result cap. Do not sweep an entity's whole note history unprompted, by paging OR by raising limit — one limit=100 call is the same runaway as five pages. To scope by date or author, pass filter_expr: the API filters before it pages, so each page holds only matching notes. Report what the first page holds and offer to fetch more. Args: entity_id: The ID of the company, person, or opportunity entity_type: What type of entity this ID refers to. Options: 0 = person, 1 = company, 8 = opportunity Take the type from the tool that produced the id (search_persons or get_person_info = 0, search_companies_* or get_company…
-
get_entities_attached_to_noteFind a page of entities directly attached to a note. For the reverse (finding notes attached to an entity), use get_notes_for_entity. A note may have more attachments than one page holds. To answer questions that depend on every attachment (counts, "is X attached", totals), keep calling with `pagination.nextCursor` until it is absent, or pass total_count=True first to learn how many there are. Args: note_id: The note ID that will be used in the search entity_type: Type of entities to retrieve. Options: 0 = person, 1 = company, 8 = opportunity. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: Number of entities to include in the page. Default is 20, max is 100. total_count: If True, include the total number of attached entities of this type as `pagination.totalCount`. Costs an extra count query, so request it only when the count itself matters. Retur…
-
search_notesSearch notes by keyword. Scope the search by passing either `note_ids` or `company_id`, or omit both to search across all notes in the organization. Each result contains a matched note and a single representative excerpt. Even if a note contains multiple matching passages, it appears exactly once in the response. IMPORTANT: `note_ids` and `company_id` are mutually exclusive. They cannot be used together. Args: prompt: The search query. Returns up to `limit` notes ordered by relevance. Prompts with no strong matches may still return low-relevance results. 3-500 characters. note_ids: Optional list of note IDs to restrict the search to. Maximum 100 ids. Mutually exclusive with `company_id`. company_id: Optional company ID to restrict the search to notes associated with that company. Mutually exclusive with `note_ids`. limit: Number of matches to return. Default is 20. Minimum is 1, maximum is 100. Returns: Structured data with a `data` array. Each item contains the matching note id, its…
-
get_entity_filesGet files attached to a company, person, or opportunity, or every file in the org. Returns metadata only (id, name, size, timestamps); file contents are not downloaded. Pass both entity_id and entity_type to scope the result to one entity. Omit both to get every file in the org, ordered reverse chronologically (newest entity_files.id first). NO SERVER-SIDE UPLOADER OR DATE FILTER: this endpoint does not accept an uploader or date-range filter. For requests like "what files did I add in the last 2 weeks", omit entity_id/entity_type, page through the org-wide results with page_token, and filter the returned entries yourself by uploader_id and created_at. An org-wide call returns ONE page of files. The org may have more files than one page holds. To answer questions that depend on every file (counts, "did I upload X", date-window sweeps), keep calling with "next_page_token" until it is absent. Args: entity_id: The ID of the company, person, or opportunity whose files should be retrieved.…
-
get_entity_file_metadataGet metadata for a single entity file (attachment / document) by ID. Use this to hydrate a file found via search_files with details it does not carry, such as who uploaded it and when. search_files only returns id, name, pageNumber, and preview; this fills in uploader_id and created_at (plus any other metadata fields returned for the file). IMPORTANT: - Use search_files (or get_entity_files) first to find a file and obtain its id; this tool does not search. - Use download_entity_file for the file's actual content; this tool only returns metadata. - This endpoint has no 404 path. A file id that is missing, deleted, or owned by another org fails validation upstream and returns a 422 whose message reads "expected <id> to be a valid id for model Affinity::Models::EntityFile". Treat that 422 as "file does not exist or is not accessible": do not retry, and do not treat it as a malformed request. An ACL denial on a file that does exist returns 403. Either way the call fails outright, never w…
-
search_filesSearch files by keyword. Scope the search by passing either `file_ids` or `company_id`, or omit both to search across all files in the organization. Each result contains a matched file and a single representative excerpt. Even if a file contains multiple matching passages, it appears exactly once in the response. IMPORTANT: `file_ids` and `company_id` are mutually exclusive. They cannot be used together. Args: prompt: The search query. Returns up to `limit` files ordered by relevance. Prompts with no strong matches may still return low-relevance results. 3-500 characters. file_ids: Optional list of file IDs to restrict the search to. Maximum 100 ids. Mutually exclusive with `company_id`. company_id: Optional company ID to restrict the search to files associated with that company. Mutually exclusive with `file_ids`. limit: Number of matches to return. Default is 20. Minimum is 1, maximum is 100. Returns: Structured data with a `data` array. Each item contains the matching file (id, nam…
-
get_opportunities_infoGet opportunities (deals) by id: name, list, and whether each is restricted. Use this when another tool returned opportunity ids without names, for example get_reminders, get_entity_files or get_entity_file_metadata. To find a deal by name, call search_opportunities. Returns no field values. Args: opportunity_ids: 1 to 100 opportunity ids. Returns: `opportunities` holds every requested opportunity the API returned. When `isRedacted` is true, the opportunity is restricted and the user is not on its allow list. Its name shows as "[Hidden]". Tell the user it exists but is hidden, and do not guess its name. `missingIds` lists requested ids that returned nothing, for example a deleted id or one from another organization.
-
search_opportunitiesSearch for opportunities or list them with optional filters. Returns one page of results. To answer questions that depend on every matching opportunity (counts, "is X in the pipeline", exports), keep calling with "next_page_token" until it is absent. Args: term: Search term used to search through opportunities page_size: Number of results per page. If not specified, defaults to 10 when searching with a term, otherwise 100. page_token: Token for the next page. Pass the value of "next_page_token" from the previous response. Do NOT parse or modify the token string. Returns: dict[str, Any]: API response containing one page of opportunities, plus "next_page_token" (pass directly to the page_token param) when more results exist. A missing "next_page_token" means this is the last page.
-
search_personsSearch persons across the whole org using filters, sorts, and/or a keyword. Omitting `search_criteria` returns every person with default pagination. WARNING: This can match many persons with large field data. To minimize response size, start with the default limit=10 (increase only if needed), use pagination (cursor) to iterate, and return only the fields you need (see FIELD SELECTION). FIELD SELECTION (which fields to return per person): `field_ids` and `field_types` are mutually exclusive; passing both raises ValueError. Valid field_types for persons are `enriched`, `global`, and `relationship-intelligence`. When both are omitted, only the person's identity (id, name, emails) is returned and `fields` is empty. Request field_types/field_ids only when needed. DISCOVERING FIELDS: Use get_entity_fields(entity_type=EntityType.PERSON, includes=[FieldInclude.FILTERABILITY, FieldInclude.SORTABILITY]) to find field IDs, valueTypes, supported operators, and any required attributeId. See Filte…
-
get_person_infoReturns basic information and non-list specific field data on the requested person. By default, returns all field data including current organization, job title, and more. Use get_entity_fields tool if you need to filter to specific fields. IMPORTANT: field_ids and field_types are mutually exclusive. They cannot be used together. Args: person_id (required): The ID of the person field_ids: list of field IDs to return. field_types: List of field types to return. Defaults to ['enriched', 'global', 'relationship-intelligence'] Returns: Metadata about the specified person including field data. A field on an opportunity you can't manage may come back with type "hidden" and a masked value.
-
get_persons_infoReturns basic information and non-list specific field data on multiple requested persons in a single call. Use this instead of calling get_person_info once per person when you already know the exact person IDs you need, e.g. contacts surfaced by get_person_relationships or get_company_relationships. Unlike get_person_info, this does not return all field data by default. Use get_entity_fields to discover field IDs/types, then pass them via field_ids or field_types. IMPORTANT: field_ids and field_types are mutually exclusive. They cannot be used together. Args: person_ids (required): The IDs of the persons. Best for up to ~100 per call (the endpoint's page size cap); use cursor for more. field_ids: list of field IDs to return. field_types: List of field types to return. Unlike get_person_info, this has no default: field data is empty for every person unless field_ids or field_types is passed. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pag…
-
get_person_list_entriesFetches list entries for the given person. Args: person_id (required): The ID of the person cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, max is 100 Returns: A dict with `data` (list of list entries) and `pagination` with `nextCursor` / `prevCursor` strings (pass directly to the `cursor` param; absent when there is no such page) and `nextUrl` / `prevUrl` full URLs (null when there is no such page).
-
get_person_relationshipsGet the relationship strengths between a person and the people they interact with (relationship intelligence / "how do I know them"). Works for both internal team members and external contacts: - Pass an internal person's ID to see their relationships with external contacts. - Pass an external person's ID to see which internal team members know them. Results are always ordered by strength (highest first), then by id, unless overridden with `order_by`. Relationships with a strength of 0 and no LinkedIn connection, and externals hidden in your org, are excluded. Each result's `interactionScore` is a float between 0 and 1. Convert it to a percentage between 0 and 100, rounded to the nearest integer (.5 or greater rounds up, less than .5 rounds down) when presenting it to the user. A relationship derived purely from a LinkedIn connection, with no interaction history, has an interactionScore of 0. Each result also has a `linkedIn` field: `{"connectedOn": <date>}` when the two people are co…
-
get_company_relationshipsGet the relationship strengths between your team and the people associated with a company ("who do we know at this company"). Returns person-to-person relationships, not a single company-level score: for the given company, each result pairs an internal team member (person1) with an external contact who works at that company (person2), plus the strength of their relationship. Results are always ordered by strength (highest first), then by id, unless overridden with `order_by`. Relationships with a strength of 0 and no LinkedIn connection, and externals hidden in your org, are excluded. Each result's `interactionScore` is a float between 0 and 1. Convert it to a percentage between 0 and 100, rounded to the nearest integer (.5 or greater rounds up, less than .5 rounds down) when presenting it to the user. A relationship derived purely from a LinkedIn connection, with no interaction history, has an interactionScore of 0. Each result also has a `linkedIn` field: `{"connectedOn": <date>}` w…
-
get_remindersReturns reminders that meet the query parameters if provided. By default, returns reminders for all entities. Results are ordered by due date, earliest first. IMPORTANT: entity_id and entity_type are only applied when both are provided; if only one is set, the entity filter is ignored and reminders for all entities are returned. Returns one page of results. To answer questions that depend on every matching reminder (counts, "is there a reminder for X", overdue sweeps), keep calling with `pagination.nextCursor` until it is absent. Args: entity_id: The ID of the company, external person, or opportunity to filter reminders by. entity_type: The type of entity that entity_id refers to. Options: 0 = person, 1 = company, 8 = opportunity. creator_id: Filters to reminders created by this internal person (identified by their person ID). owner_id: Filters to reminders assigned to this internal person (identified by their person ID). completer_id: Filters to reminders completed by this internal p…
-
get_saved_viewsPaginate through all saved views the user has access to for a specific List. Args: list_id (required): ID of the list. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 100, maximum is 100. Returns: Data array of saved views (each containing id, name, type, createdAt) plus a pagination object with `nextCursor` / `prevCursor` strings (pass directly to the `cursor` param; absent when there is no such page) and `nextUrl` / `prevUrl` full URLs (null when there is no such page).
-
get_saved_view_list_entriesPaginate through list entries on a given saved view. Each list entry contains basic information about the related person/company/opportunity along with the field data that the saved view has been configured to display. Unlike search_list_entries, field selection is not supported here: the saved view itself defines which fields (columns) are returned. Use get_saved_views to discover the view_id for a list. Args: list_id (required): ID of the list. view_id (required): ID of the saved view on that list. cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: The number of items to include in the page. Default is 10 to keep response size manageable, maximum is 100. Page through with cursor for more entries. Returns: Data array of list entries (entity info plus the view's configured fields) plus a pagination object with `nextCursor` / `prevCursor` strings (pass…
-
get_transcript_fragmentsGet a page of dialogue fragments (individual speaker turns) from a transcript. Fragments are free text and a long meeting has thousands of them, so a full transcript does not fit one tool result. Do not sweep a whole transcript unprompted — by paging or by raising limit. Report what the first page holds and offer to fetch more. To get transcripts for a specific entity (company, person, opportunity): 1. Use get_notes_for_entity to get notes for that entity 2. Look for notes where type is "ai-notetaker" or "ai-notetaker-reply" 3. Those notes have a transcriptId field - use it with this tool Args: transcript_id: The transcript ID from the transcriptId field in AI Notetaker notes (returned by get_notes_for_entity). cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pagination.prevCursor` from a prior response. Do NOT parse or modify the cursor string. limit: Number of items per page. Default is 20, max is 100. Response size scales with how much eac…
-
get_usersGet a page of users (internal team members) in your Affinity organization. More users may exist than one page holds: keep calling with `pagination.nextCursor` until it is absent to see everyone. Use this to list, search, or paginate through internal Affinity users, not external contacts. To look up only the authenticated caller, use get_current_user instead. Args: term: Case-insensitive search across first name, last name, and primary email address. Example: "jane" or "[email protected]". filter_expr: Filter expression using Affinity query language. Supported fields: - id: unique identifier for the user (operators: =) Example: "id=1" or "(id=1 | id=2)" - status: the user's account status (operators: =). Values: active, invited, deactivated Example: "status=active" Boolean logic: & (AND), | (OR), () for grouping Full filtering spec: https://developer.affinity.co/pages/external-api-v2/filtering cursor: Cursor for the next or previous page. Pass the value of `pagination.nextCursor` or `pa…
-
create_companyCreate a new company in Affinity. IMPORTANT: Call `search_companies_top_matches` first with the company's name or domain to avoid creating duplicates. Affinity does not deduplicate on create. Args: name: The company's name. Must be at least 1 character. Example: "Acme Corporation" domain: The company's primary domain, hostname only, no protocol or path. Example: "acme.com" person_ids: IDs of persons to associate with the company at creation time. Use `search_persons` to look up person IDs first. These associations may be applied asynchronously by the API. Example: [1234, 5678] Returns: The newly created Company object including id, name, domain, domains, and global fields. If `person_ids` were provided, the returned company may still show an empty or stale `persons` list immediately after creation because person associations are assigned asynchronously. Fetch the company again later if you need to verify the associated persons.
-
send_feedbackSends feedback to Affinity about MCP server improvements or additions. IMPORTANT: Use this tool when you cannot fulfill the user's request due to limitations in the current toolset. You MUST immediately offer to submit feedback about the missing capability. Use HTML formatting for the body. Args: subject: Short summary of the feedback. body: The feedback itself, HTML formatted: what was requested and which capability was missing. Returns: No content if the request was successful
-
create_fieldCreate a new field (custom column) on a list or globally on an entity. IMPORTANT: - value_type takes a string such as "text" or "dropdown-multi". A "-multi" type holds many values. The returned id is a field_id such as "field-1234", not a number. - This tool does NOT create dropdown options. When value_type is a dropdown or ranked-dropdown, the field is created with no options. For a list-specific field (list_id provided), pass the returned id as field_id to `create_list_field_dropdown_option`. For a global field (list_id omitted), options must be managed in the Affinity app. - Omitting `list_id` creates a global field on every entity of `entity_type`. This requires org-level "manage global fields" permission and will fail without it. Prefer passing a `list_id` unless the user explicitly wants a global field. - If the user asks for a "dropdown" field without saying whether it should allow one selection or multiple, ask before calling this tool. "dropdown" holds one selection, "dropdow…
-
create_list_field_dropdown_optionCreate a new dropdown option on a list's dropdown field. Adds a selectable value to a dropdown, ranked-dropdown, or status-dropdown column on a list, e.g. a new "Not Relevant" option on a Communication Status field or a new "Negotiating" status on a deal pipeline's Status field. IMPORTANT: - option_type must match the field's actual type. A mismatch returns a 400. Call `get_list_field_dropdown_options` first and use the `type` of the existing options. Do not rely on `get_list_fields`: it reports a list's Status field as valueType "ranked-dropdown", but its options are "status-dropdown". - For a ranked-dropdown field, both rank and color are required. For a plain dropdown field, none of rank, color, status_category, or win_rate may be set. - For a status-dropdown field (a list's Status field), rank, color, and status_category are all required; win_rate is optional. Args: list_id: ID of the list the field belongs to. Use `get_lists` to look it up. field_id: Human-readable field ID, e.g.…
-
create_interactionLog an interaction (meeting, call, or chat message) with one or more people. Args: interaction_type: The kind of interaction to log: - InteractionType.MEETING (0) - InteractionType.CALL (1) - InteractionType.CHAT_MESSAGE (2) If unclear which type the user means, ask before calling. date: When the interaction occurred, in ISO 8601 format. Example: "2025-01-31T10:56:29Z" person_ids: IDs of the persons involved in the interaction. At least one is required. Use search_persons to look up person IDs by name or email. content: Notes describing the interaction. direction: Direction of the message. Required when interaction_type is CHAT_MESSAGE; not applicable to meetings and calls (omit it for those). - ChatMessageDirection.SENT (0): message sent to the contact - ChatMessageDirection.RECEIVED (1): message received from the contact Returns: dict[str, Any]: API response containing the created interaction.
-
create_listCreate a new list for companies, people, or opportunities. IMPORTANT: The entity type is fixed when the list is created and cannot be changed later. A company list holds companies, a person list holds people, and an opportunity list holds opportunities. IMPORTANT: Setting is_public=True requires the "Share accessible Lists globally" permission. Without it, the request fails with a 403. Default to a private list (is_public=False) unless the user explicitly asks for a public/shared list. This tool only creates the empty list. To add entities, call create_list_entry (companies/people) or create_opportunity (opportunities) afterward. Args: name (required): Name of the list (1 to 255 characters). entity_type (required): Entity type the list holds. One of EntityType.COMPANY, EntityType.PERSON, or EntityType.OPPORTUNITY. is_public: Whether the list is visible to everyone in the organization. Defaults to False (private to the creator). Returns: JSON object with the new list's id, name, type,…
-
create_list_entryAdds an existing company or person as a list entry on the specified list. Opportunities CANNOT be added using this tool. The entity must match the list type: a company id for a company list, a person id for a person list. A mismatch returns 400. An opportunity list returns 400. This tool cannot set field values, so a list with a required field returns 400 and names the missing field. If the V2 list entry write feature gate is off for the caller, the call returns 404. Args: list_id: ID of the list where the entry is being added. entity_id: ID of the existing person or company to add to this list. Returns: JSON object containing the newly created list entry: id, type, listId, creatorId, createdAt, and entity (the company or person it holds).
-
create_noteCreate a note attached to entities or a meeting, or as a reply to an existing note. IMPORTANT: When picking `meeting_id` from get_meetings_for_entity, ALWAYS use the response's `v2_id` field if it is present and non-null. Only fall back to the `id` field when `v2_id` is absent. Meetings, calls, and chat messages each allow at most one root note. This tool can create that root note only for a meeting, using meeting_id. A meeting synced from a calendar may not have one yet, so creating a meeting note when one already exists fails with an error that names the existing note's id. Calls and chat messages are created with their note, so they always already have a root note. To add to a call or chat message, or to a meeting that already has a note, reply instead: set note_type=USER_REPLY and parent_note_id to that note's id, and tell the user you are replying. parent_note_id accepts the root note of a meeting, call, or chat message. To find the note id: - create_interaction and get_meetings_…
-
create_opportunityCreates a new Opportunity (deal) on the specified list. The list must be an opportunity-type list. Companies and/or persons can optionally be associated with the new Opportunity at creation time. The list entry on the target list is auto-created, no separate `create_list_entry` call is needed. IMPORTANT: Call `search_opportunities` first with the opportunity's name to avoid creating duplicates. The name must be unique on the list. NOTE: Custom field values cannot be set during creation. If the user wants to set field values on the new Opportunity, follow up with `get_list_fields` to discover field IDs and `upsert_list_entry_field_values` with the returned `listEntryId`. Restricted opportunities: only users on the allow list can see a restricted Opportunity. Pass restricted=true only when the user asks for a restricted, private or confidential deal. Some organizations allow only restricted opportunities, some allow only open ones, and some do not have the feature. If this tool says the…
-
create_personCreates a new person (contact) in Affinity. Use for adding people not yet in the CRM, e.g. importing from a CSV or a hand-entered contact. IMPORTANT: - Call `search_persons` first with the person's name or email to avoid creating duplicates. Affinity rejects an email address that already exists, but do not rely on it to catch other duplicates, such as the same name with a different email. - `company_ids` takes Affinity company IDs (integers), not company names or domains. If the user provides a company name or domain, resolve it via `search_companies_top_matches` first and pass the resulting IDs. If unclear, ask before calling. Args: first_name: Required. The person's first name. last_name: Required. The person's last name. emails: Required. List of email addresses for the person. Pass an empty list `[]` if none are known. At most 100 addresses. Example: `["[email protected]"]`. company_ids: Optional list of Affinity company IDs the person is associated with. Example: `[1687449]`. The…
-
create_reminderCreates a new reminder. PRIVACY: The reminder gets the default access setting of the organization. In an organization with private defaults, only people with access can read it. SCHEDULING BY DAY NAME: If the user specifies a recurring reminder by day name (e.g. "every Friday", "every Sunday"), calculate the next occurrence of that day from today's date and use it as the due_date for the first reminder. Set reminder_days to match the recurrence interval (e.g. 7 for weekly, 14 for bi-weekly). You must still supply reset_type and the other recurring fields documented under reset_type and reminder_days below; this section only helps derive due_date and reminder_days. Args: owner_id: Required. Internal person ID who owns the reminder (team member only; not an external contact). reminder_type: Required. ReminderType.ONE_TIME (0) or ReminderType.RECURRING (1). If unclear, ask before calling. entity_id: Required with entity_type. ID of the company, external/collaborator person, or opportunit…
-
update_companyUpdate an existing company (organization) in Affinity. IMPORTANT: Global companies (enriched from Affinity's global dataset) cannot have their name or domain changed. Only manually created (non-global) companies accept name/domain edits. Call `search_companies_top_matches` or `get_company_info` first if you are unsure whether the company is global. At least one of `name`, `domain`, or `person_ids` must be provided; a ValueError is raised before any API call otherwise. Args: company_id (required): The ID of the company to update. name: New name for the company. Omit to leave the name unchanged. Example: "Acme Corporation" domain: New primary domain, hostname only, no protocol or path. Omit to leave the domain unchanged. Example: "acme.com" person_ids: Full set of persons associated with the company. This replaces the existing associations, it is not additive, so include every person who should remain associated. Use `search_persons` to look up IDs first. Omit to leave the associations…
-
delete_companyPermanently deletes one or more companies (organizations) from Affinity by company id. Confirmation gate: with confirm=false (the default) nothing is deleted. The response is a summary (warning, companies_preview, total, global_count, related_totals, next_step). Show the whole warning to the user in your own message, unquoted, keeping every company name it shows, the count, and the sentence that says what is kept: a short question alone hides how far the deletion reaches. Call again with confirm=true only after the user approves (or already approved this deletion in the conversation). A user who wants to see what the deletion affects does not need another delete_company call: read it yourself with get_company_list_entries, get_company_info and get_reminders. next_step carries the details. WARNING: this deletes the company record itself, not just a list membership. List entries on all lists, field values, and tagged reminders are deleted with it. Notes, files, people, and opportunities…
-
merge_companiesMerges two companies (organizations): the duplicate's data moves onto the primary, which survives, unless a global company forces a new combined record. Confirmation gate: with confirm=false (the default) nothing is merged and the response is a preview (primary_company, duplicate_company, outcome, warning, next_step). Show the whole warning to the user in your own message, unquoted, naming both companies and which one survives. Call again with confirm=true only after the user explicitly approves this specific merge; one approval per merge, never a batch. Direction: primary_company_id survives and duplicate_company_id disappears (deleted, or hidden when it is a global company), except for outcome "new_company" (a global primary gaining a domain, or two globals), where Affinity replaces BOTH records with a new company that takes the primary's name and both records' domains; find it afterwards with search_companies_top_matches by one of those domains. If it is unclear which record the us…
-
update_fieldUpdate a field (custom column): its name, description, or required flag. Works on a list-specific field, or on a global company field. Pass only the values you want to change. IMPORTANT: - Use the field's `type` from `get_list_fields` or `get_entity_fields`. For "list", pass the list_id. For a "global" or "enriched" company field, omit list_id. - Person fields and global opportunity fields are not supported. A field on an opportunity list is a list field, so pass its list_id. - Only custom fields can be renamed or get a description. Enriched and native company fields (for example "affinity-data-location") accept only is_required. - A list field needs the Standard role on the list. The list's Status and Owners fields, and any is_required change, need the Admin role on the list. A company field needs the org-level "manage global fields" permission. - A name that another field already uses returns a 400 ("<name> is an existing column."). - The id that `create_field` returns is the field_…
-
delete_fieldPermanently delete a field (custom column) and every value set on it. Works on a custom field of a list, or on a custom global company field. Confirmation gate: with confirm=false (the default) nothing is deleted. The response is a summary (warning, field, records_with_values, list_name for a list field, next_step). records_with_values is the number of list entries or companies with a value on the field, or null when Affinity could not count them. Show the whole warning to the user in your own message, keeping the field name, the list name and the count. Call again with confirm=true only after the user approves (or already approved this deletion in the conversation). IMPORTANT: - Use the field's `type` from `get_list_fields` or `get_entity_fields`. For "list", pass the list_id. For a "global" company field, omit list_id. - Person fields, global opportunity fields, and enriched or native fields are not supported. A field on an opportunity list is a list field, so pass its list_id. - A…
-
update_list_field_dropdown_optionUpdate (rename or restyle) a dropdown option on a list's dropdown field. Use this to rename a dropdown option (e.g. rename "Pass" to "Passed"), to change a ranked-dropdown or status-dropdown option's rank or color, or to change a status-dropdown option's status category or win rate. Pass only the values you want to change. IMPORTANT: - rank and color only apply to ranked-dropdown and status-dropdown fields. Setting them on a plain dropdown field returns a 400. - status_category and win_rate only apply to status-dropdown fields (a list's Status field). - Call `get_list_field_dropdown_options` first to get the dropdown_option_id. Args: list_id: ID of the list the field belongs to. field_id: Human-readable field ID, e.g. "field-1234". dropdown_option_id: ID of the dropdown option to update. text: New label for the option (1 to 255 characters). Must stay unique on the field. rank: New sort position (0 to 2147483647). Ranked-dropdown and status-dropdown only. color: New tag color. Options:…
-
delete_list_field_dropdown_optionDelete a dropdown option from a list's dropdown field, e.g. removing an outdated "Associate" option from a Seniority field. IMPORTANT: - Deleting an option cascades: it is removed from every list entry that had it selected. This cannot be undone. - Works on dropdown, ranked-dropdown, and status-dropdown fields. - Call `get_list_field_dropdown_options` first to get the dropdown_option_id. Args: list_id: ID of the list the field belongs to. field_id: Human-readable field ID, e.g. "field-1234". dropdown_option_id: ID of the dropdown option to delete. Returns: None. The API returns 204 No Content on success.
-
delete_list_entriesDeletes one or more list entries, removing those entities' rows from a list. Confirmation gate: with confirm=false (the default) nothing is deleted — the response is a summary of what would be deleted (warning, entries_preview, total, list_type, list_name, next_step). Present the warning to the user in your own message, unquoted, keeping every entry name it shows, the count, the list name, and the sentence that says what the deletion does to the entity itself. A short confirmation question is not enough on its own: the user cannot tell how far the deletion reaches. Call again with confirm=true only after the user explicitly approves (or already approved this specific deletion in the conversation). For deletions spanning several lists, collect each list's summary first and ask once. This tool works on company and person lists only. There only the membership is removed; the entity is untouched. It CANNOT remove entries of an opportunity list: the API refuses them. To delete an opportuni…
-
upsert_entity_field_valuesUpsert up to 100 field values for a company or person. `upserts` is ALWAYS a JSON list, even for a single field. Wrap every item in `[ ... ]`. Passing a bare object instead of a one-item list is rejected ("Input should be a valid list"). Use this tool when the entity is NOT on a list. If the entity is on a list and you have list context, prefer `upsert_list_entry_field_values`. Opportunities are not supported. Opportunity fields live on list entries. Use `upsert_list_entry_field_values` instead. Each item's `value.type` must be EXACTLY one of the allowed type strings and must match the field's declared type. Copy it verbatim, don't retype it. Args: entity_type: 0 = person, 1 = company. entity_id: ID of the company or person. upserts: A LIST of field value items. Each item is {"id": "<field-id>", "value": {"type": "<field-type>", "data": <...>}}. Use `get_entity_fields` to discover field IDs and their types, and `get_entity_field_dropdown_options` for dropdown option IDs. See `ValueToU…
-
upsert_list_entry_field_valuesUpsert up to 100 field values for a single list entry (one PATCH). `upserts` is ALWAYS a JSON list, even for a single field. Wrap every item in `[ ... ]`. Passing a bare object instead of a one-item list is rejected ("Input should be a valid list"). Use this for fields on a list/pipeline entry, including opportunity fields. For company or person fields when the entity is NOT on a list, use `upsert_entity_field_values`. Before calling: - Get field IDs and each field's value type from `get_list_fields`. Field IDs are list-specific (`field-XXXXX`); don't reuse IDs from another list or entity. - Each item's `value.type` must be EXACTLY one of the allowed type strings and must match the field's declared type. Copy it verbatim, don't retype it (e.g. `ranked-dropdown`, not `ranked-dropdwon`). - For dropdown, ranked-dropdown, or dropdown-multi fields, get option IDs from `get_list_field_dropdown_options`. Args: list_id: ID of the list. list_entry_id: ID of the entry on this list. upserts: A L…
-
update_noteEdit an existing note: replace its text, or change which persons, companies, or opportunities it is attached to. Use this to fix a typo or a wrong association instead of creating a second note. IMPORTANT: only the note's creator can update it; any other caller gets a 403. The content of a note that contains @mentions cannot be changed (400). Reply notes (user-reply, ai-notetaker-reply) accept content only: passing person_ids, company_ids, or opportunity_ids for a reply is rejected. A note's type cannot be changed. Do not retry a 403 or 400 with the same arguments. IMPORTANT: person_ids, company_ids, and opportunity_ids REPLACE the existing associations of that kind, they do not append. Omit a list to leave those associations unchanged. Pass an empty list ([]) to clear that kind, as long as at least one association of any kind remains: a root note cannot be left with none, and clearing the last one fails with 400 "Cannot remove all associations". To add or remove a single association,…
-
delete_notePermanently deletes one or more notes from Affinity by note id. Confirmation gate: with confirm=false (the default) nothing is deleted. The response is a summary (warning, total, next_step). Show the whole warning to the user in your own message, unquoted, keeping the count and the sentence that says what is kept: a short question alone hides how far the deletion reaches. Call again with confirm=true only after the user approves (or already approved this deletion in the conversation). A user who wants to see what the deletion affects does not need another delete_note call: read it yourself with query_notes and get_entities_attached_to_note. next_step carries the details. WARNING: the note's attached files go with it. Deleting a root note also deletes every reply under it, user and AI Notetaker replies alike; deleting a reply removes only that reply. Pass the root only: a reply listed in the same call as its root is already gone by the time its own DELETE runs and comes back as a 404 u…
-
update_opportunityUpdates an existing Opportunity: renames it or changes its company/person associations. IMPORTANT: company_ids and person_ids REPLACE the existing associations, they do not append. Passing company_ids=[5] on an Opportunity already linked to companies 5 and 9 removes the link to 9. To add or remove a single association, first fetch the current set (e.g. via `search_opportunities` or `get_list_entries`) and pass the full desired list. Passing an empty list (company_ids=[]) clears all of that association type. Omitting a param leaves it unchanged. Args: opportunity_id: ID of the Opportunity to update. name: New name for the Opportunity. Omit to leave the name unchanged. company_ids: Full set of company IDs to associate. Replaces existing companies. Omit to leave the associations unchanged; [] clears all. person_ids: Full set of person IDs to associate. Replaces existing persons. Omit to leave the associations unchanged; [] clears all. Returns: JSON object containing the updated Opportuni…
-
delete_opportunityPermanently deletes one or more opportunities (deals) from Affinity by opportunity id. Confirmation gate: with confirm=false (the default) nothing is deleted. The response is a summary of what would be deleted (warning, opportunities_preview, total, next_step). Present the warning to the user in your own message, unquoted, keeping every opportunity name it shows and the count. A short confirmation question is not enough on its own: the user must see which opportunities the call deletes. Call again with confirm=true only after the user explicitly approves (or already approved this specific deletion in the conversation). WARNING: this deletes the opportunity record itself, not just a list membership. There is no undo. If the user only wants to rename or re-associate an opportunity, use update_opportunity. If the user removes rows from a company or person list and you hold list entry ids, use delete_list_entries instead. delete_list_entries cannot remove opportunity list entries: to remo…
-
update_personUpdates an existing person (contact) in Affinity: renames them or changes their email addresses or company associations. IMPORTANT: `emails` and `company_ids` REPLACE the existing values, they do not append. Passing `company_ids=[5]` on a person already linked to companies 5 and 9 removes the link to 9. To add or remove a single value, first fetch the current set (e.g. via `get_person_info`) and pass the full desired list. Passing an empty list (`emails=[]`) clears all of that value. Omitting a param leaves it unchanged. `company_ids` takes Affinity company IDs (integers), not company names or domains. If the user provides a company name or domain, resolve it via `search_companies_top_matches` first and pass the resulting IDs. If unclear, ask before calling. At least one of `first_name`, `last_name`, `emails`, or `company_ids` must be provided; a ValueError is raised before any API call otherwise. Names and emails are saved first, then companies, in two separate API calls. If the comp…
-
delete_personPermanently deletes one or more people (contacts) from Affinity by person id. Confirmation gate: with confirm=false (the default) nothing is deleted. The response is a summary (warning, persons_preview, total, internal_count, related_totals, next_step). Show the whole warning to the user in your own message, unquoted, keeping every person name it shows, the count, and the sentence that says what is kept: a short question alone hides how far the deletion reaches. Call again with confirm=true only after the user approves (or already approved this deletion in the conversation). A user who wants to see what the deletion affects does not need another delete_person call: read it yourself with get_person_list_entries, get_person_info, get_reminders and query_notes. next_step carries the details. WARNING: this deletes the person record itself, not a list membership and not a company link. List entries on all lists, field values, reminders tagged to the person, and notes the person wrote are d…
-
merge_personsMerges a duplicate person into a primary person: the duplicate's data moves onto the primary, which survives, and the duplicate is deleted. Use when the user asks to merge, combine, or de-duplicate two people (contact records). IMPORTANT: - This tool has no undo. An admin can undo the merge later from Settings > Duplicate Management > Merged. All of the duplicate person's data (notes, list entries, relationships, and more) moves to the primary person, then the duplicate is deleted. - Confirmation gate: confirm=false (the default) merges nothing and returns a preview (primary_person, duplicate_person, warning, next_step). Show the whole warning to the user in your own message, unquoted, naming both persons and which one survives. Call again with confirm=true only after the user explicitly approves this specific merge; one approval per merge, never a batch. Both calls raise a 404 when either person is not found: the API would accept that merge and fail it later. - duplicate_person_id ca…
-
update_reminderUpdates an existing reminder. IMPORTANT: - The tool changes only the fields that you give. Other fields stay the same. - Only the creator or the owner of the reminder can update it. Other users get 403 or 404. - You cannot change the type of a reminder. To convert a reminder, call delete_reminder, then call create_reminder. - reset_type and reminder_days apply only to a recurring reminder. The API returns 400 for a one-time reminder. You can give one of them, or both. - If you give no field to change, the tool raises ValueError. Args: reminder_id: The ID of the reminder to update. owner_id: Internal user ID to reassign the reminder to. Find IDs with get_users. content: Description or note text for the reminder. due_date: ISO 8601 timestamp at noon UTC. Omit it to keep the existing due date. Use noon UTC (e.g. "2025-01-31T12:00:00Z"). The API stores due_date as an instant and renders it in the user's local timezone; noon UTC lands on the correct calendar day in every timezone from UTC-…
-
delete_reminderPermanently delete a reminder by ID. There is no undo. Only the creator or the owner of the reminder can delete it. Other users get a permissions error. A reminder that does not exist, or that the current user cannot read, returns a 404. Args: reminder_id: The ID of the reminder to delete. Returns: {"success": true, "reminder_id": reminder_id}. A 404 means the reminder is missing or hidden from the current user.
-
suggest_field_updatesRun the Data Update Agent on a note or meeting to get suggested field updates (SFUs). The agent reads the note or meeting and suggests new values for fields on the companies, persons and opportunities it mentions. Nothing is written until the user applies the suggestions. Find the note or meeting first with the note and meeting tools (search_notes, get_notes_for_entity, get_meetings_for_entity). If more than one matches the user's words, show the candidates and ask. Never guess. The tool starts generation, then checks it every 5 seconds for up to 20 seconds, and returns within 25 seconds. A meeting note runs as its meeting. While a run for the subject is still generating, a repeat call waits on that run and starts no second one. Otherwise each call starts a new run, and the user's previous open set for the subject becomes done: its undecided suggestions can no longer be decided. Args: note_id: The note id. Pass exactly one of note_id, meeting_id or note_url. meeting_id: The meeting id…
-
apply_suggested_field_updatesApply suggested field updates (SFUs): accept, edit or reject each pending suggestion in a set, and write the accepted values to the records. Confirmation gate: with confirm=false (the default) nothing is written. The response is a preview: "summary", one "rows" item per pending suggestion (decision, record, field, current_value, new_value), "warning" and "next_step". Show the summary and every row in your own message. Call again with the same arguments and confirm=true only after the user approves. IMPORTANT: Every pending suggestion needs a decision. A suggestion in none of accept, reject or edits raises ValueError naming it, before any write. Ask the user about those; never decide for them. An id in edits is accepted with that value. reject and edits win over accept_all. Call get_suggested_field_updates first for the suggestion ids, fields and value types. Args: suggested_field_update_id (required): The set id, a UUID. accept: Suggestion ids to accept as suggested. accept_all: true…
-
dismiss_suggested_field_updatesDismiss suggested field updates (SFUs): reject every pending suggestion in a set and close it. Nothing is written to the records. Confirmation gate: with confirm=false (the default) nothing changes. The response has "total", a "warning" with the fields, and "next_step". Show the warning. Call again with confirm=true only after the user approves. To keep some suggestions, use apply_suggested_field_updates instead. Args: suggested_field_update_id (required): The set id, a UUID. confirm: false returns the preview only; true dismisses. Returns: confirm=true: {"summary", "status"}, for example "Rejected 3. The set is done." Errors: ValueError when the set is not open. 404 when the set is missing or not this user's, or the user has no access to suggested field updates.
-
issue_entity_files_upload_tokenGet a single-use upload target for attaching a file to a company, person, or opportunity. Returns a PUT target and a single-use, time-limited token — it does NOT upload the file itself. The caller must PUT the raw file bytes to the returned url with the returned headers, within expires_in seconds. IMPORTANT: - This tool never sees or returns file bytes. If you can make HTTP requests yourself, PUT the file directly using the returned method/url/headers. If you cannot, give the user `curl_example` to run themselves. - The Authorization header in the response carries a single-use token, not the caller's own API key. Treat it as a secret — it authorizes exactly one upload and expires after expires_in seconds either way. - Accepted file extensions/MIME types and a 100 MB size limit are enforced both here and when the PUT is redeemed; a mismatch here fails fast, but passing validation here does not guarantee the PUT will succeed if the declared metadata doesn't match the file that's actuall…
-
issue_entity_files_download_tokenGet a single-use download target for a file attached to a company, person, or opportunity. Returns a GET target and a single-use, time-limited token — it does NOT return the file itself. The caller must GET the returned url with the returned headers, within expires_in seconds. IMPORTANT: - This tool never returns file bytes. If you can make HTTP requests yourself, GET the file directly using the returned method/url/headers. If you cannot, give the user `curl_example` to run themselves. - The Authorization header in the response carries a single-use token, not the caller's own API key. Treat it as a secret — it authorizes exactly one download and expires after expires_in seconds either way. Args: file_id: The unique ID of the entity file to download. Use get_entity_files to list an entity's files and obtain a file_id. Returns: dict with: - method: always "GET" - url: the endpoint to GET the file from - expires_in: seconds until the token expires - headers: headers to include on the GET…
-
company_signalsCollect recent dated public signals (product launches, funding rounds, executive hires and departures, acquisitions, partnerships, IPO filings, press coverage, earnings) for a given list of companies, each with a source link and a confidence state. Use whenever the user asks "what's new at these companies?", "any recent news, funding, or exec changes for X, Y, Z?", "what have these companies launched lately?", or wants a news scan of named companies or of an Affinity list (a list link works). Other skills also call it to get signals for their own company lists. Searches company blogs and feeds first, then the web. Read-only: never writes to the CRM. Always use this skill for company news scans instead of answering from general knowledge.
-
data_migrationUse when a user wants to import, migrate, or load data into Affinity — companies, people, or opportunities — from any source: another CRM, a CSV or spreadsheet export. Also use when the user wants to re-import records that failed a previous import, or clean and prepare a file for Affinity's CSV importer. Do NOT use for: LinkedIn connections import (Settings → LinkedIn Data), email/meeting interaction history (comes from email sync), or restoring deleted data.
-
data_quality_auditAudit the health of an Affinity CRM and deliver a workbook and a one-page HTML audit summary: what is in place today (Your CRM), every check that was run and what it found (Checks), one row per finding with what to do, who does it and a status to track (Findings), and the reference tabs behind them. Use when a user asks for a "data quality audit" or a "CRM audit", to "audit our CRM", "audit our data", "check data quality", "how healthy is our Affinity data", "clean up our CRM", "data hygiene report", or asks what is wrong with or missing from their lists, fields, companies or people. Also trigger for a single audit by name ("just audit our lists"). Also trigger when an audit request asks to find or merge duplicates: the skill explains that duplicate records are out of scope and points to Duplicate Management. Read-only and recommendation-only: never modifies records and never executes clean-up, even when asked.
-
event_setupSet up a complete event tracking workflow in Affinity CRM for private capital teams. Use this skill whenever a user wants to track an event in Affinity — including creating global fields for attendance, building lists for attendees (companies and/or people), adding list-level fields for event-specific data, sourcing attendees from CRM relationships, reviewing historical event data, drafting event communications, or processing post-event attendance from a CSV or manual list. Also use when the user mentions "event list", "conference tracking", "attendee list", "event setup in Affinity", "post-event follow-up", or "update attendance status". This skill runs an interactive, end-to-end workflow directly against the Affinity MCP.
-
founder_referencesFind reference-check sources on a named person (a founder, CEO, executive, hiring candidate, operating partner, or advisor), or on the founders of a named company. Builds the background from a pitch deck or other material the user has, public sources such as funding announcements, and Affinity files, then uses Affinity relationship data to find who on the user's team knows them, former colleagues and ex-employees, early investors and co-investors, angels and board members, customers that are portfolio or pipeline companies, people seen in meetings with them, and what past notes already say. Use whenever the user asks for references, backchannel, reference checks, "who can vouch for", "who worked with", "who can tell me about" a person, or wants diligence on a founder or management team, or is vetting a candidate for a portfolio company, operating partner, board, or advisor role, even if they never say the word "reference". Read-only: never writes to the CRM or sends anything.
-
lp_pipeline_briefReport what changed in an LP fundraising pipeline in Affinity over a chosen window — new and increased commitments, stage changes, losses, and LPs needing attention — grounded only in Affinity data: notes, logged field changes, list-entry creation dates, meetings, and last-email dates. Use whenever the user asks what changed in their LP pipeline, wants an LP, fundraising, or investor-relations update, asks for a weekly brief, asks who moved or what happened with LPs or limited partners since last week, or wants fundraising progress from Affinity. Trigger even without the word "brief": "what's new with our LPs", "catch me up on the raise", and "anything move on the fundraise" all belong here. For LPs investing into the firm's funds, not for deal or portfolio pipelines. Reads the CRM; writes only on explicit confirmation.
-
market_map_generatorGenerates an interactive investment thesis market map by searching Affinity for companies matching a thesis, scoring each for thesis fit, and rendering a scatter plot (funding stage on the X axis, engagement depth on the Y axis) plus coverage tables. Use this skill whenever a user asks to "build a market map", "generate a market map", "map out my thesis", "what companies do we have in [space]", "show me our coverage of [sector]", or pastes a thesis and asks to visualize it against their Affinity data. Also trigger for "what does our pipeline look like for [thesis]".
-
meeting_prepProduce a relationship-intelligence pre-meeting brief from Affinity CRM data via the Affinity MCP. Use whenever the user asks to "prep me for", "prepare for", "brief me on", "get me ready for", or "tell me about" an upcoming meeting, call, person, or company, or asks what they should know before a meeting. Always runs a full multi-tool Affinity sequence and synthesizes it, never a single lookup. Read-only: never writes to the CRM.
-
passed_deal_reviewAnti-portfolio report: finds companies the firm passed on or lost, while there was still time to invest, that went on to raise their next-stage round in a chosen period. Ranks the most painful misses, shows where every data point comes from, and breaks the misses down by pass reason, industry, geography and team. Use when someone asks for an anti-portfolio review or report, missed deals, "who did we pass on that just raised", a quarterly or annual pass review, deals they personally passed on, or patterns in pass reasons. Runs on demand against one deal pipeline list in Affinity.
-
portfolio_check_inRanks portfolio companies by whether they need attention right now — specifically, companies with no recent internal touchpoint (meeting, note, email) but a recent external signal (funding, exec change, press, competitor move). Surfaces the company-news mismatch case first (news that names the portco), separates competitor news (news about its competitors or sector), and distinguishes "couldn't check" from "checked, nothing found." Use when asked for a portfolio check-in, "who needs attention," or a portco monitoring pass.
-
ready_to_raise_indicatorFlags which companies in an Affinity list or pipeline are statistically due to raise again, using the Affinity MCP connector -- no custom artifact or hosted tool needed, just Claude plus the connector. Trigger on typing "RTR" (case-insensitive, standalone or as a prefix like "RTR my pipeline list") or on a plain-language ask like "which of my portfolio companies are due to raise," "check this list for re-raise signals," "run the ready-to-raise estimate," or pasting/uploading/referencing a company list with a raise-timing ask. Runs the full methodology directly against the user's live Affinity data, calculated fresh at the moment it's invoked.
-
warm_introSurface ranked warm introduction paths to a target person or company from Affinity's intro paths endpoint, which ranks the paths itself. Use this skill whenever a user asks how to get introduced to someone, who can introduce them to a contact, how to reach a target at a specific firm, or any variation of: "warm intro", "intro path", "who knows X", "can you find me an intro to Y", "how do I get connected to Z", "who can introduce me to someone at [Company]", or "who in my network knows [Person]". Also trigger if the user asks to draft an intro request or to see the network map again after paths are surfaced. Always use this skill for warm intro questions — do not attempt to answer from general knowledge or guess at relationships. Read-only; it never re-ranks, reorders, or filters the paths it is given.
Connect your AI to Affinity CRM
Add the official Affinity CRM MCP server to Claude, ChatGPT or your editor.
Connect Affinity CRM MCP to Claude
Claude.ai and Claude Desktop
Open Customize, then Connectors.
Click +, then Add custom connector.
Name it Affinity CRM, paste the address below and click Add.
Connect Affinity CRM MCP to ChatGPT
Open Settings, then Apps, then Advanced settings.
Turn on Developer mode, then click Create.
Name it Affinity CRM and paste the address below.
ChatGPT usually only connects to servers with self-serve sign-in or none.
Connect Affinity CRM MCP to Cursor
Opens Cursor if it's installed. If nothing happens, use the JSON below.
Connect Affinity CRM MCP to VS Code
Opens VS Code if it's installed. Otherwise add the JSON settings from the Cursor tab to your .vscode/mcp.json.
Affinity CRM MCP server questions
Does Affinity CRM have an official MCP server?
Yes. Affinity CRM publishes its MCP server itself. Its address is on Affinity CRM's own domain (mcp.affinity.co). MartechDB only lists servers a vendor publishes, never community copies.
What is the Affinity CRM MCP server URL?
The address is https://mcp.affinity.co/mcp.
Does the Affinity CRM MCP server need an API key?
No, but it needs set-up first: it signs in with OAuth through an app you (or an admin) create in your Affinity CRM account.
How do I connect Affinity CRM MCP to Claude, ChatGPT or Cursor?
In Claude, add https://mcp.affinity.co/mcp as a custom connector (Customize, Connectors, Add custom connector). In Cursor, use the Add to Cursor link on MartechDB's page, or paste the JSON settings into your mcp.json. In Claude Code, run: claude mcp add --transport http affinity-crm https://mcp.affinity.co/mcp
What can the Affinity CRM MCP server do?
It offers 97 tools, including get_current_user, search_companies_top_matches, search_all_companies, get_company_info and get_company_list_entries (as listed on 22 Nov 2026). 62 are marked read-only and 22 can delete or overwrite data.
Is the Affinity CRM MCP server in the official MCP Registry?
No. Affinity CRM hasn't published it to the official MCP Registry.
Is the Affinity CRM MCP server free to use?
No, it needs a paid plan. Plan needed: Scale, Advanced, and Enterprise plans. It costs more than the base plan. From Affinity CRM's own pages, checked on 7 Oct 2026.
Is the Affinity CRM MCP server working?
Yes. It answered our most recent check. We last checked it on 22 Nov 2026.
What are the alternatives to Affinity CRM's MCP server?
In CRM, the most agent-friendly alternatives with an official MCP server are Attio (97/100), Zero (97/100), Pipedrive (96/100), NetHunt (95/100) and Twenty (94/100). Affinity CRM scores 75/100 on the same index.
Alternatives to Affinity CRM's MCP server
Vendors in CRM with an official MCP server, the most agent-friendly first. Affinity CRM scores 75/100.
Compare all 55 CRM vendors →Change history
What we've seen change for Affinity CRM, newest first.
-
Affinity CRM: Affinity CRM MCP server: 1 tool added (now 97)
Use with AI
Connect your AI to MartechDB
Then ask it about Affinity CRM, or any of the 1,850 marketing tools with an official MCP server. Free, no sign-in.
Claude.ai and Claude Desktop
Open Customize, then Connectors.
Click +, then Add custom connector.
Name it MartechDB, paste the address below and click Add. There's no sign-in.
The free plan allows one custom connector. On Team and Enterprise plans, an owner adds the connector first.
Claude Code
Run this in your terminal:
ChatGPT
Open Settings, then Apps, then Advanced settings.
Turn on Developer mode, then click Create.
Name it MartechDB, paste the address below and choose No authentication.
Developer mode is on Plus and Pro plans, and on Business, Enterprise and Edu if an admin allows it.
API
The same data as JSON, with no sign-in. See the OpenAPI description for every endpoint and field.
CLI
Search the directory and print the command to connect a vendor's server, from your terminal. Needs Node 18 or later. View on npm
Show that your MCP server is verified
If you work at Affinity CRM, add this badge to your MCP docs or README. It's free, it's the same for every vendor, and it updates from our daily checks: it shows uptime while the server responds, and says so if it stops.
Something wrong or out of date? Suggest a correction or email us, and we'll check it against the evidence. Every correction we make is listed on Corrections.