Skip to main content
Connect your coding agent to Beam to deploy projects and manage your workspace:
Restart your agent after setup. The local server runs with beam mcp and uses your selected CLI context. See agent setup for client options.

​
Local Tools

Give the agent your project’s absolute path. Python projects also need the SDK installed locally.

​
Workspace Tools

whoami shows the current workspace and its credit status. Use name for the newest active deployment or deployment_id for a specific version. See deployment lifecycle for restarts and rollbacks. update_config accepts paths such as runtime.cpu, runtime.memory, and autoscaler.min_containers. CPU is in millicores (1000 = 1 CPU); memory is in MB. Use set_env or connect_services for environment references.
Configuration changes deploy a new version. Secret updates take effect on the next container start. Database credential rotation restarts the database and connected apps.

​
Confirmation and Scope

App, deployment, secret, and database deletion require confirm: true, as do non-GET calls through api. Use api_routes to list routes and api to call them. {ws} expands to your workspace ID. Workspace permissions apply.

​
Remote Clients

Use https://app.beam.cloud/api/v1/mcp with Authorization: Bearer <token> in your client configuration. It accepts JSON-RPC POST requests, without a GET event stream. Local projects require beam mcp. For documentation search, use https://docs.beam.cloud/mcp.