# MCP command

Find the aggregate or per-edge MCP endpoint.

Source: https://railgrid.ai/docs/reference/cli/mcp/


Use `kubectl railgrid mcp url --mcpserver-name default` for the workspace aggregate, or `kubectl railgrid mcp url --edge EDGE-NAME` for one Kubernetes edge. The `--edge` form does not support server/Linux edges; use the aggregate endpoint for Linux/SSH tools. Exactly one flag is required.

See [connect external AI assistants through MCP](https://railgrid.ai/docs/use/ai-assistants/) for setup, authentication, and verification.