Skip to main content
Any AI agent can pull live user activity. Wiring one is always the same three moves — only the per-agent specifics change.

​
1. 🔌 Connect via MCP (recommended)

The recommended way for an agent to read live activity is the MCP server — one endpoint, the get_live_user_activity tool, Bearer auth. Point your agent at it and it can pull a user’s recent activity on demand.
Optional fallback — REST. If an agent can’t speak MCP but can call an external HTTP endpoint (e.g. Intercom’s “data connectors”), the same activity is available at the REST endpoint with the same Bearer auth. Reach for this only when MCP isn’t an option.

​
2. 🔑 Satisfy identity

Make the agent send the same user_id that activity is stored under — your activity source’s stable user id (e.g. the PostHog identify id, the Amplitude user_id). This is the step that most often gets skipped — see Identity.

​
3. 🤖 Follow the per-agent recipe

Intercom Fin

Connect Fin to the MCP server + Messenger JWT identity verification.

More agents coming

Maven and others follow the same pattern — pick a door, bind identity to your source’s user id, point the agent at the endpoint.
Adding a new agent doesn’t change this surface — the endpoint, the MCP tool, and the identity rule stay the same. A new agent just needs its own short recipe describing how it calls the endpoint and how it passes a verified user_id.