Personal agents

Connect Pi to MCP Chats

One block of settings gives Pi your processes, the data and email templates you've allowed, and a way to wait for your approval.

Setup

Get your address and an access token
In your workspace, open Agents and create an access token. The page shows your exact MCP address and fills it, and the token, into these settings for you. Tokens are shown once, so keep yours somewhere safe.
2
Pi has no MCP support built in. Add it with the pi-mcp-adapter extension, then restart Pi.
Command
```
pi install npm:pi-mcp-adapter
```

Add this to ~/.config/mcp/mcp.json, or .mcp.json in a project, and set PROCESS_FLOW_TOKEN to your access token.

Configuration

{
  "mcpServers": {
    "process-flow": {
      "url": "https://<your-mcp-endpoint>/mcp",
      "auth": "bearer",
      "bearerTokenEnv": "PROCESS_FLOW_TOKEN"
    }
  }
}

Check it's connected
Ask Pi: “What processes can you see?” It should list the published processes in your workspace. If it can't reach the server, check the address and that the token hasn't been revoked.

The adapter connects when Pi first needs the server, and reaches all its tools through one proxy tool.

pi-mcp-adapter on GitHub

This setup is written from Pi's own MCP documentation.

Bring the routine you keep explaining.

Join the waitlist for our free pilot. When invited, we’ll help you turn one recurring job into a process and try it with your agent. For solo users and teams, with a setup call to get you started.

Join the waitlist