firecrawl

Firecrawl MCP Server provides a configurable interface to run Firecrawl's web scraping capabilities via the Model Context Protocol (MCP). The server runs a pre-packaged MCP client that communicates with Firecrawl's cloud or self-hosted API to perform crawling, discovery, search, and content extraction tasks. You can drive it from MCP-enabled environments (like Cursor, Windsurf, or Claude) by configuring an MCP server block that points to the Firecrawl MCP entry point and supplying your API key. The server supports automatic retries, rate limiting, and both cloud and self-hosted deployment modes, making it suitable for batch scraping, large-scale discovery, and targeted data collection. Tools such as streaming via SSE are supported, allowing real-time results to be consumed by your application.

To use the server, start it via npx with your API key and then connect via MCP tooling. For Cursor, Windsurf, or VS Code integrations, configure the MCP server block with the proper command and environment variable, then reference the server in your MCP workflows or prompts to perform scraping tasks like site discovery, content extraction, or batch re-scraping with retry logic.

Prerequisites:

• Node.js and npm/yarn installed on the host
• An API key from Firecrawl (FIRECRAWL_API_KEY)

1. Install the MCP server globally (via npm):

npm install -g firecrawl-mcp

2. Run the MCP server (providing your API key):

env FIRECRAWL_API_KEY=your-api-key npx -y firecrawl-mcp

3. Alternatively, run with debugging or custom environment settings (optional):

env FIRECRAWL_API_KEY=your-api-key DEBUG=true npm run firecrawl-mcp

4. If you are integrating with Cursor, Windsurf, or other MCP-enabled environments, use the appropriate configuration block described in the README to point the MCP client at this server. For example, in Cursor you would add the provided JSON snippet to enable the MCP server pointing to npx -y firecrawl-mcp with FIRECRAWL_API_KEY set.

5. Confirm the server is reachable at the standard MCP endpoint (e.g., http://localhost:3000/mcp) depending on your transport mode.

Notes and tips:

• Always keep FIRECRAWL_API_KEY secure; do not commit it to public repositories.
• If you switch to a self-hosted Firecrawl API, set FIRECRAWL_API_URL to the endpoint of your instance.
• You can customize retry and rate-limiting behavior using the environment variables documented in the Firecrawl Configuration section (e.g., FIRECRAWL_RETRY_MAX_ATTEMPTS, FIRECRAWL_RETRY_INITIAL_DELAY).
• For local testing with a UI or integration, you can use the Streamable HTTP Local Mode by prefixing the run command with HTTP_STREAMABLE_SERVER=true to route traffic via HTTP locally.
• In MCP configurations for Cursor/Windsurf/VS Code, ensure the API key placeholder is replaced with a secure reference or prompt, as shown in the README examples.
• The npm package name is firecrawl-mcp and the recommended command to run is npx -y firecrawl-mcp, which will fetch and start the MCP client packaged with Firecrawl support.