MCP Integration

claude mcp add --transport http cranl https://app.cranl.com/api/mcp

CranL runs a hosted Model Context Protocol (MCP) server, so AI coding assistants can deploy apps, read logs, create databases, run sandboxes and more.

URL

https://app.cranl.com/api/mcp (Streamable HTTP)

Sign-in

OAuth: the client opens a browser window and you approve. Or send Authorization: Bearer cranl_sk_….

Tools

The 174 CranL operations, named after the operation (for example deploy_app). See MCP Tools.

There’s no binary to install and no key to paste: clients register through OAuth dynamically.

Supported clients

  • Claude Code

  • Cursor

  • VS Code

  • Windsurf

  • Clients that only support stdio, through cranl mcp

  • Any client that supports remote MCP servers

Quick start

  1. Add the server to your client (see Client Setup), or run cranl init and let it set up Claude Code, Cursor, VS Code and Windsurf for you.

  2. Approve the sign-in in the browser window that opens.

  3. Ask your assistant something like “List my CranL apps” or “Deploy this folder to germany-1”.

How it works

  • Tools are CranL operations. A default set is loaded; find_tools and call_operation reach the rest. You can choose toolsets in the URL.

  • Progress: long calls such as deploys and sandbox creation report progress while they run.

  • Images: screenshots and charts come back to the model as images.

  • Docs: the CranL docs are available as cranl://docs/{topic} resources.

  • Safety: destructive tools ask for a confirmation first, and the agent can request the owner’s approval. See Safety Controls.

Security

Every request needs an OAuth token or an API key. A key’s scopes decide which tools can run. Use ?readonly=1 or a Read only key to stop the assistant from changing anything. Rate limit: 300 requests per 60 seconds per key or token.