AI + MCP

Connect your AI to Deploy Hatch.

Give GitHub Copilot, Claude, or another compatible MCP client the ability to inspect, deploy, diagnose, and recover your application through Deploy Hatch.

Create a project

Connect the GitHub repository your AI will build and deploy.

Create an agent credential

Use a project-scoped credential with only the permissions the agent needs.

Connect your AI

Point the MCP client at https://deployhatch.com/api/mcp and authenticate with the credential.

GitHub Copilot

Connect Deploy Hatch to Copilot in VS Code with a project-scoped credential.

Claude

Connect Claude Code to Deploy Hatch over remote MCP.

Other MCP clients

Use any client that supports remote HTTP MCP and bearer authentication.

No project UUID required

A project-scoped agent credential identifies the project the agent is authorized to operate on. The AI can discover its credential context instead of asking you for Deploy Hatch internal IDs.

Recovery stays bounded

Diagnostics tell the agent whether source, configuration, platform recovery, or human input is required. Agents must stop at human-action boundaries and never invent secrets.

What the autonomous flow looks like

Your AI
  ↓
Deploy Hatch MCP
  ↓
Credential + project context
  ↓
Inspect workload → Deploy → Diagnose → Authorized recovery
  ↓
running + healthy
  ↓
Public application URL