Affinity MCP Reference

​

What is Affinity MCP?

Affinity MCP (Model Context Protocol) is the server interface that lets external AI tools query and update your Affinity CRM. It’s a standards-based way for AI clients to call structured tools — search_companies, create_note, get_opportunity — against live Affinity data, with your user’s permissions enforced on every call. For setup steps, follow the per-client guides at developer.affinity.co/pages/mcp/supported-clients.

​

Who should use Affinity MCP

• Associates & analysts doing deal research, IC prep, founder meeting prep
• Partners running portfolio check-ins and LP communication prep
• Operators keeping the CRM clean: updating fields, creating notes, setting reminders from chat
• Developers building agentic workflows over Affinity data

​

When to use MCP vs Affinity AI Chat

See Affinity AI Chat Overview for the in-app option.

​

Use cases by role

​

Associates and analysts — deal research and IC prep

• “Find seed and Series A SaaS companies in healthcare we’ve never contacted.”
• “Summarize all notes on Acme Corp from the last 90 days.”
• “Show me upcoming meetings with founders this week.”
• “Who on our team has the strongest relationship with the partners at Benchmark?”
• “Create a note on Acme Corp: [paste meeting summary]“

​

Partners — portfolio and LP context

• “List portfolio companies I haven’t checked in with this quarter.”
• “Summarize last three interactions with [LP firm name].”
• “Show me deals where the next milestone is overdue.”

​

Operators — CRM hygiene

• “Show me opportunities with no Stage field set in the last 30 days.”
• “Update the Stage field on the Acme deal to ‘Due Diligence’.”
• “Set a reminder to follow up with Jane Smith in two weeks.”

​

Developers — automation

• Use MCP from a custom-built agent (e.g., LangChain, your own orchestrator) to enrich, sync, or summarize.
• See Available tools for the full schema.

​

Hosted vs Local

For most users: pick Hosted. For Local setup, see Local Setup at developer.affinity.co.

​

What data MCP can access

MCP tool calls execute as the authenticating user. You can read and update anything you’d be able to read or update directly in Affinity — and nothing else.

• List access: only lists you’ve been granted access to
• Field visibility: respects field-level permissions
• Role gates: enterprise role restrictions still apply
• OAuth read-only mode: at consent time, OAuth users can uncheck the write scope to block all create / update calls through MCP

​

Capabilities and limitations

​

What MCP can do

• Query companies, people, opportunities, lists, fields, notes, meetings, reminders, relationship strength
• Create notes and reminders
• Update field values
• Search by tag, list membership, recency, owner, status, and more
• Return data filtered by your user’s CRM permissions

​

Current limitations

• MCP cannot create lists or fields (use the Affinity UI or API)
• MCP cannot bulk-update entries across many lists in one call (work entry-by-entry, or use the API)
• MCP can’t see file attachments, only their metadata
• Some clients (Copilot CLI / Studio, Gemini CLI, Notion, ChatGPT) don’t currently support OAuth — those use API keys instead

​

Authentication

Two methods, both managed under Affinity → Settings → Manage Apps:

• OAuth: One-click browser auth. Supported by Claude (all surfaces), ChatGPT (Plus+), Notion (Custom Agents).
• API key: A long-lived key passed as a Bearer token in the Authorization header. Supported everywhere. Rotate or revoke at any time.

See Authentication at developer.affinity.co for full details.

​

Security and privacy

• The MCP server is stateless. No CRM data, prompts, or AI responses are persisted.
• Affinity sees only the structured tool calls your AI client makes — not your typed prompts. Your prompts are governed by your AI provider’s policy.
• All tool calls run with your user’s CRM permissions.
• OAuth tokens are user-scoped and revocable. API keys can be rotated.

Full details at developer.affinity.co/pages/mcp/security.