A self-hosted Rust crawler that gives Claude direct access to scrape, crawl, extract, map, and search operations through a Firecrawl-compatible API. The MCP server runs embedded with no external dependencies, idles at roughly 50 MB RAM, and exposes the same /v1/scrape and /v1/crawl endpoints you'd hit over REST. Reach for this when you need structured data extraction from websites during agent workflows without spinning up Node.js or browser sidecars. It ships as a single static binary you can run via npx crw-mcp for zero-install MCP integration, or point at api.fastcrw.com if you want managed infrastructure. The AGPL license means you can self-host free or pay for the hosted version to avoid license obligations.
claude mcp add --transport stdio us-crw uvx crw