# crawl4ai-mcp-server

This repository is a pinned, tested distribution of the upstream Crawl4AI web-crawling toolkit, packaging its native MCP server behind Docker Compose with setup and deployment guidance rather than shipping a custom MCP implementation.

It runs the official upstream Crawl4AI container at a pinned, locally verified version and exposes its native Model Context Protocol server, reachable over server-sent events or WebSocket once an API token is generated for the local shell. A smoke-test script starts an isolated stack, checks public health, confirms that unauthenticated MCP access is refused and that authenticated schema access works, and then tears the stack down again, and a separate script validates the Compose configuration statically without starting anything.

Because Crawl4AI 0.9.1 only supports the legacy server-sent-events and WebSocket MCP transports rather than the newer single-endpoint Streamable HTTP transport, an MCP client needs legacy transport support, such as Claude Code's own SSE transport option; Claude Desktop's cloud-hosted custom connectors cannot reach a loopback-only deployment like this one.

It is packaged with Shell scripts and Docker Compose, and released under the MIT license.

Website: https://github.com/BjornMelin/crawl4ai-mcp-server