API and MCP
An API and an MCP server for everything
The dashboard is a client of the public API, so anything you click you can script. AI agents get the same tools through MCP, never more than their key allows.
Every action is an endpoint
Sites, environments, domains, backups, and billing all have API endpoints, described in an OpenAPI file.
Scoped API keys
Name each key and give it scopes, an expiry, and an IP allowlist. The secret is shown once.
MCP server
Connect Claude Desktop or another MCP client with an API key and let it read and change your sites.
Confirm guards
Tools that delete, reset, deploy, or roll back need a confirmation, so a vague prompt cannot destroy a site.
Jobs you can follow
Every server change runs as a job with events. Follow it with server-sent events or polling.
Audited like people
Agent actions appear in the activity log with the key that made them.
Built for scripts
- Problem Details errors with a request ID to quote
- Idempotency keys on every write
- 600 requests a minute per key, with rate limit headers
- Cursor pagination on every list
Read more in the docs
Start with one site. Add more when you need them.
Sign in with Google or GitHub, choose a plan, and your first site is on staging in minutes.