MCP overview
MCP (Model Context Protocol) lets an assistant call tools against your real data instead of guessing at it. Envpilot runs a remote MCP server, so Claude, Claude Code, Codex or Cursor can work with your projects' variables, secret files and shared accounts, with the same permissions, rate limits and audit trail as the dashboard.
Requires the Pro plan (mcp_server). Connect by signing in with your Envpilot login. The old key-based setup lives on API keys (legacy).
https://www.envpilot.dev/api/mcpTransport is Streamable HTTP. Auth is an OAuth access token from the Envpilot sign-in.
What an agent can do#
| Can | Cannot |
|---|---|
| List the projects you can see | Delete anything |
| Read variables, secret files and shared accounts you can read | Read outside your role's projects or environments |
| Create variables, files and accounts; update variables; replace files | Write a protected environment without approval |
| Search project names and variable keys | See variable values through search |
| Read and write documentation drafts | Publish a page |
The agent does what your role allows, nothing more. A write into a protected environment becomes a change request that a second person approves, and documentation stays a draft until a person publishes it.
Getting connected#
- Add the server to Claude, Claude Code, Codex, Cursor or VS Code, with no
Authorizationheader. - Sign in when your browser opens.
- Confirm with
/mcpthat the tools loaded, not merely that the server is listed.
Limits#
- Pro plan only, re-checked on every call — a downgrade stops the agent immediately.
- A connection acts as you. Your role decides what it sees; there is no way yet to give it less.
- Rate limits apply per person: 120/min metadata, 30/min value pulls, 30/min variable and account writes, 20/min file uploads. See Tools.
- Every value-returning call and every write is audited against your user.