Verified

Firecrawl MCP Server

Scrape and search live web data for AI agents with structured output formats.

Data last scanned yesterday · Reviewed yesterday

83 1 dangerous tool26 tools★ 7,028JavaScriptMIT

Overview

This MCP server enables AI agents to access, scrape, and search the live web by converting web content into clean, structured data ready for AI consumption. It supports complex operations including web page interaction, navigation, and deep research tasks through an autonomous agent interface. The tool offers both a fully featured endpoint requiring an API key and a search-only, read-only surface. It is compatible with various deployment methods, including direct integration via npx or configuration within environments like Cursor.

Our verdict

With a security score of 53/100, this server is classified as high risk. Because the project is community-maintained and has not yet undergone editorial review, users should exercise caution regarding its integration into sensitive environments. It is best suited for developers building research-heavy agents who can manage their own API key usage and understand the security implications of third-party web interaction tools.

Supports both cloud and self-hosted deployment models
Small supply-chain surface with only four direct dependencies
Provides a dedicated read-only search surface
Frequent updates with latest commit 0 days ago

Community-maintained rather than official vendor support
Authentication and read-only modes lack formal editorial review

Setup

Claude Desktop

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

Paste into ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows), then restart Claude Desktop. Some servers require API keys or tokens as environment variables — check the project's README and add an "env" object if needed.

No read-only mode — running this grants full read/write access.
No built-in authentication — anyone with access to this config can invoke every tool it exposes.

Tools

Tool	Description	Risk
firecrawl_scrape	Scrape content from a single URL with advanced options.	low
firecrawl_map	Map a website to discover all indexed URLs on the site.	low
firecrawl_search	Search the web and optionally extract content from search results.	low
firecrawl_search_feedback	Sends structured feedback on a previous firecrawl_search result.	low
firecrawl_feedback	Sends structured feedback for a completed v2 endpoint job.	low
firecrawl_crawl	Starts a crawl job, polls until it reaches a terminal state, and returns the final crawl status/data.	low
firecrawl_check_crawl_status	Check the status and results of an existing crawl job by ID.	low
firecrawl_parse	Parse local files or hosted upload references with Firecrawl's /v2/parse endpoint.	low
firecrawl_extract	Extract structured information from web pages using LLM capabilities.	low
firecrawl_agent	Autonomous web research agent that independently browses, searches, navigates, and extracts data.	low

Compatibility

Client	Local	Docker	Remote	Read-only
Claude Desktop
Cursor
VS Code
Windsurf
ChatGPT

Frequently asked questions

›Can I use the Firecrawl MCP server without an API key?

Yes, you can use the hosted keyless free tier, which allows limited access to the scrape, search, and interact tools. However, an API key is required to access the full tool set, including crawl, map, agent, and extract, as well as to increase rate limits.

›What is the difference between the main hosted endpoint and the search-only endpoint?

The main endpoint provides access to all tools, while the search-only endpoint is a read-only surface limited to six specific tools, including firecrawl_search and the firecrawl_research series. The search-only endpoint does not perform page-content fetching.

›Does the server support local execution?

Yes, you can run the server locally using npx, or via a streamable HTTP mode by setting the environment variable HTTP_STREAMABLE_SERVER=true.

›Are there specific configuration requirements for running this on Windows?

If you encounter issues while configuring the server on Windows, you should prepend your command with cmd /c "set FIRECRAWL_API_KEY=your-api-key && npx -y firecrawl-mcp".