Using AI Agents
Let Claude, ChatGPT, Cursor or your own scripts manage your CRM.
PersonalCRM is built so an AI assistant can operate your whole CRM: write notes from a sentence, prep you before a meeting, run bulk migrations, dedup contacts, "create reminders for everyone I met at the conference", and anything else that is tedious by hand.
There are two ways to give an agent access. For almost everyone, the first is the right one.
The easy way: MCP
PersonalCRM ships a built-in MCP server with 18 tools. Claude, ChatGPT, Claude Code, Cursor, VS Code and any other MCP client connect by signing in with your account: no keys to copy, no schema to explain.
- In the app, open Settings → MCP.
- Pick your client and follow the two or three steps (Cursor and VS Code have one-click installers).
- Talk to your assistant.
Things people actually ask:
- "Who am I overdue to catch up with? Draft a short message for the top three."
- "Log a call with Jane and note that she started at Vercel."
- "Move everyone from my University network who now works in tech into the Work network."
- "Find contacts I haven't spoken to in over 6 months and create reminders for next week."
The MCP docs cover the endpoint, authentication details and the full tool list.
For scripts and custom agents: the REST API
If you are building your own agent or automation, use the REST API instead. Give your agent two things:
- An API key: create one named after the agent in Settings → API (treat it like a password: it grants full access to your data).
- This instruction:
PersonalCRM API: base URL
https://personalcrm.me/api/v1, authenticate every request withAuthorization: Bearer <key>. Fetchhttps://personalcrm.me/api/openapi.jsonfor the full schema.
That is all an agent needs. The OpenAPI spec describes every endpoint, parameter and response shape in machine-readable form.
Good practice
- One key per agent so you can revoke access individually.
- Delete keys you no longer use: revocation is immediate.
- MCP apps you connected are listed under Settings → MCP → Connected apps, where access can be revoked at any time.
- Agents can be told to work read-only ("only GET requests"), but keys and grants have full access (scoped access is on the roadmap).