An MCP server for AI agents#
Hangar now has a hosted MCP server, so the agent that wrote your code can also ship it. From Claude Code, Cursor, VS Code or any client that speaks the Model Context Protocol, the agent can:
- find the service your repository deploys to, deploy it and follow the build;
- read build logs, runtime logs, metrics, requests and traces;
- change variables, domains and build settings, and wire databases to apps;
- create projects, apps from GitHub and databases;
- deploy local code that is not in Git, through an upload link that works once.
You sign in with OAuth: your browser opens Hangar, you pick the organization the agent acts in and approve it. The agent never has more access than you, its calls are marked in the audit log, and it is rate limited. See Hangar for agents and the MCP docs.
Pre-deploy command#
A pre-deploy command runs once before each deployment goes live, in the image that was just built. It is the place for database migrations, such as npx prisma migrate deploy. If it fails, the deployment stops and the running version stays. Set it in Settings → Runtime, or from an agent with configure_service.
A smoother first deploy#
- Connect GitHub in a popup without leaving the onboarding, then follow the first build from there.
- The onboarding asks for the port your app listens on, and every app gets a
PORTvariable that matches its domain. - Domains route to the app's real port, and Hangar checks that the app answers.
Isolation#
- Each environment runs on a private network of its own.
- Every Docker Compose service gets a CPU, memory and process limit.
