docs / mcp

Connect greyout to your AI agent

greyout exposes its monitoring API as a Model Context Protocol (MCP) server so any MCP-aware agent — Claude Code, Cursor, Aider, custom — can manage projects, check incidents, and mint cron-monitor endpoints without copy-pasting tokens.

endpoint

https://mcp.greyout.io/mcp

Streamable HTTP transport, JSON-RPC 2.0, Bearer auth. Discovery is automatic via the standard OAuth Protected Resource metadata document at /.well-known/oauth-protected-resource.

how the auth flow works

  1. Your agent posts to /mcp without a token.
  2. It receives 401 with a resource_metadata pointer.
  3. It auto-registers (RFC 7591) at our Authorization Server.
  4. It opens a browser to /oauth/authorize; you log in (or sign up via magic link) and pick the workspace + scopes.
  5. The agent exchanges the code for tokens (PKCE-protected) and starts using tool calls.

tools exposed

  • greyout_whoami
  • greyout_list_projects
  • greyout_get_project_state
  • greyout_sync_project
  • greyout_delete_project
  • greyout_create_cron_endpoint
  • greyout_delete_cron_endpoint
  • greyout_list_incidents

device management

Every connected agent appears in Settings → Authorized devices with its name, granted scopes, and last-used timestamp. Revoke at any time — revocation is instant and cascades to every issued access + refresh token.

Stuck or curious? See the OAuth docs for the protocol-level details.