Firecrawl MCP Server

The Firecrawl MCP server provides a specialized, read-only search surface at the /v2/mcp-search endpoint. This surface is designed for restricted access and only exposes a specific subset of tools. It is distinct from the full MCP surface available at /v2/mcp.

Key Characteristics:

• Read-only: Only specific search and research tools are available.
• No Page Content Fetching: The firecrawl_search tool in this surface does not support scrapeOptions. It is strictly limited to search parameters to prevent fetching third-party page content.
• Strict Schema: Input schemas are strict; unknown fields are rejected.
• Authentication Required: Unlike the full surface, the keyless free-tier fallback does not apply to the search surface. Every request (including tools/list) requires valid credentials.

The Firecrawl MCP Server supports two versions of the Firecrawl API to maintain backward compatibility while providing modern features:

• V1 (Legacy): Based on Firecrawl JS 1.29.3. It includes extended tools like firecrawl_deep_research and firecrawl_generate_llmstxt but uses legacy parameter names.
• V2 (Current): Based on Firecrawl JS 3.1.0. It uses modern API parameter names (e.g., ScrapeOptions, MapOptions), enhanced JSON extraction with schema support, and improved caching defaults.

When using the scrape tool, select the output format based on your requirements:

• JSON format (Recommended): Use this when you need specific data from a page. Define a schema to extract exactly what you need. This is the best practice to keep responses small and prevent context window overflow in LLMs.
• Markdown format: Use this sparingly. Only select Markdown when you require the full page content, such as for full-article summarization or analyzing the overall page structure.

Run the MCP server immediately using npx by providing your Firecrawl API key as an environment variable.

env FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp

Add the following configuration to your ./codeium/windsurf/model_config.json file:

{
  "mcpServers": {
    "mcp-server-firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Depending on your deployment target, use the following commands to start the server:

Cloud Service (Versioned Endpoints):

npm run start:cloud

Local Development (V2 by default):

npm start

SSE Local Server:

SSE_LOCAL=true npm start

HTTP Streamable Server:

HTTP_STREAMABLE_SERVER=true npm start

To upgrade from V1 to V2, follow these steps:

1. Update Endpoint URLs:
   • Change /:apiKey/sse to /:apiKey/v2/sse
   • Change /:apiKey/messages to /:apiKey/v2/messages
2. Update Tools:
   • Remove usage of firecrawl_deep_research and firecrawl_generate_llmstxt (these are V1 specific).
   • Update extraction calls to use the new schema format to leverage enhanced JSON extraction.
3. Adjust Parameters:
   • Note that maxAge now defaults to 172800000ms (48 hours) instead of 0.
   • onlyMainContent now defaults to true.

To run the server using Streamable HTTP locally instead of the default stdio transport, use the following command and connect to the provided URL:

Command:

env HTTP_STREAMABLE_SERVER=true FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp

URL: http://localhost:3000/mcp

To set up the project for development, install the dependencies, build the project, and run the test suite using the following commands:

# Install dependencies
npm install

# Build
npm run build

# Run tests
npm test

You can connect to a remote hosted server without any local setup.

• Keyless Free Tier: Use the endpoint below for rate-limited access to scrape, search, and interact. Other tools like crawl, map, agent, and extract require an API key.
• With API Key: Use your Firecrawl API key to unlock the full toolset and higher limits.

Endpoints:

• Keyless: https://mcp.firecrawl.dev/v2/mcp
• With API Key: https://mcp.firecrawl.dev/{FIRECRAWL_API_KEY}/v2/mcp