Firecrawl MCP Server

Firecrawl MCP Server is the official Model Context Protocol server from Firecrawl. It connects MCP-compatible clients such as Cursor, Claude Desktop, Windsurf, and VS Code to Firecrawl's search, scrape, map, crawl, and interact capabilities so agents can work with live web content. It is MIT-licensed and can run against the Firecrawl cloud API or a self-hosted instance.

What It Is

The server exposes Firecrawl's web data API as MCP tools. An agent can start from a query or a known URL and get back clean markdown or JSON matching a schema it supplies. The README lists up to 27 tools in the full profile, covering scraping, URL mapping, site crawling, document parsing, web search, autonomous research, recurring page monitoring, and credit usage checks.

How the Tools Fit Together

The README gives guidance on choosing a tool:

• firecrawl_scrape for a known URL, with JSON output preferred to keep responses small.
• firecrawl_map to discover URLs on a site without fetching content.
• firecrawl_crawl for multi-page extraction, bounded by limits and depth.
• firecrawl_search when starting from a query, optionally fetching page content in the same call.
• firecrawl_interact for pages that need clicks, typing, or navigation.
• firecrawl_agent for asynchronous multi-source research that returns structured data.
• firecrawl_monitor_* for recurring checks with diffs and change alerts.

Additional tools cover developer search over GitHub issues, merged pull requests, READMEs, and docs; US government primary-law search; paper research; and Alexandria provider catalogue lookup and execution.

Setup and Deployment

There are several ways to connect. A hosted endpoint offers a keyless free tier limited to scrape, search, and parse (rate-limited), with OAuth or API-key connections unlocking the full tool set. Locally, the server runs with npx -y firecrawl-mcp, as a global npm install, in Docker (a published GHCR image), or over Streamable HTTP. The README includes configuration for Cursor, Windsurf, Claude Desktop, and VS Code. A FIRECRAWL_API_URL variable points the server at a self-hosted Firecrawl API. Automatic retries and rate limiting are listed, and feedback tools can be disabled through environment variables.

Programmatic Use

The package also has an unstable library entry point for embedding the server in another runtime, with hooks for authentication, extra tools, and HTTP routes. The README warns the API may change in any release.