Firecrawl MCP ServerFeatured

What is Firecrawl MCP Server?

Firecrawl MCP is the official Model Context Protocol server from Firecrawl that lets AI agents search the web, scrape pages into clean Markdown or JSON, crawl and map sites, and interact with live pages from Claude, Cursor, VS Code and other clients.

You can use the hosted endpoint at https://mcp.firecrawl.dev/v2/mcp with no install, or run the open-source firecrawl-mcp package locally with npx or Docker. The local server also works with a self-hosted Firecrawl instance.

How to use Firecrawl MCP Server?

Prerequisites: a Firecrawl API key from https://www.firecrawl.dev/app/api-keys for the full tool set. Node.js is only needed for the local npx install.

Hosted, keyless: point your client at https://mcp.firecrawl.dev/v2/mcp. Without a key, firecrawl_scrape, firecrawl_search and firecrawl_parse work with rate limits; crawl, map and agent tools need a key.

Hosted, signed in: use https://mcp.firecrawl.dev/v2/mcp-oauth and complete the OAuth flow in your client. This URL is an MCP endpoint, not a page to open in a browser.

Hosted, API key: keep the URL https://mcp.firecrawl.dev/v2/mcp and send the header Authorization: Bearer <FIRECRAWL_API_KEY>. The README says never to put the key in the URL or in an agent chat.

Local:

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

Claude Code

# Hosted with OAuth sign-in
claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp-oauth

# Local server
claude mcp add firecrawl -e FIRECRAWL_API_KEY=fc-YOUR_API_KEY -- npx -y firecrawl-mcp

Claude Desktop

Add to claude_desktop_config.json:

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

Cursor

Open Cursor Settings > Features > MCP Servers, click + Add new global MCP server, and enter:

{
  "mcpServers": {
    "firecrawl-mcp": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR-API-KEY"
      }
    }
  }
}

VS Code

In .vscode/mcp.json, prompting for the key instead of storing it:

{
  "inputs": [
    { "type": "promptString", "id": "apiKey", "description": "Firecrawl API Key", "password": true }
  ],
  "servers": {
    "firecrawl": {
      "command": "npx",
      "args": ["-y", "firecrawl-mcp"],
      "env": { "FIRECRAWL_API_KEY": "${input:apiKey}" }
    }
  }
}

Windsurf

Add the same mcpServers block (server name mcp-server-firecrawl) to ./codeium/windsurf/model_config.json.

For self-hosting, set FIRECRAWL_API_URL to your instance. Set HTTP_STREAMABLE_SERVER=true to serve the local server over Streamable HTTP at http://localhost:3000/mcp. A Docker image is published as ghcr.io/firecrawl/firecrawl-mcp-server.

Key features of Firecrawl MCP Server

• Web search that can return full page content, plus a search index for coding agents covering GitHub issues, merged pull requests, READMEs and docs
• Scrape any URL to Markdown, HTML, links, screenshots or JSON that matches a schema you supply
• Crawl and map whole sites with path filters, depth and page limits
• Live browser interaction: click, type and navigate before reading a page
• Asynchronous research agent, document parsing (PDF, Word, spreadsheets), and recurring page monitors with diffs
• Credit usage checks, automatic retries and rate-limit handling
• Hosted (keyless, OAuth or API key) or local via npx/Docker, including self-hosted Firecrawl

Firecrawl MCP Server tools

With the full profile and default settings, the server lists 27 tools; the keyless hosted endpoint exposes only firecrawl_scrape, firecrawl_search and firecrawl_parse.

• Fetch and parse: firecrawl_scrape reads one known URL; firecrawl_parse converts a document such as a PDF. Structured extraction is done with firecrawl_scrape and formats: ["json"] (firecrawl_extract is deprecated).
• Discover: firecrawl_search starts from a query; firecrawl_map lists a site's URLs without fetching them.
• Crawl: firecrawl_crawl collects many pages; firecrawl_check_crawl_status reads a job's progress.
• Interact: firecrawl_interact opens a live browser session; firecrawl_interact_stop closes it.
• Research: firecrawl_agent and firecrawl_agent_status run multi-source research jobs; the firecrawl_research_* tools search papers in PubMed, bioRxiv, medRxiv and arXiv.
• Monitor: firecrawl_monitor_* tools schedule repeat checks and report changes.
• Account and feedback: firecrawl_credit_usage, firecrawl_search_feedback and firecrawl_feedback.

Use cases of Firecrawl MCP Server

• Pulling current docs into a coding session: "Scrape the Next.js caching docs and summarize what changed."
• Research with sources: "Search for recent benchmarks of vector databases and give me the top five with links."
• Structured extraction: "Get the plan names and prices from this pricing page as JSON."
• Site audits: "Map example.com and list every page under /blog."
• Pages behind interactions: "Open this page, click Load more, then read the results."
• Change tracking: set a monitor on a competitor's pricing page and get alerted when it changes.

Firecrawl MCP Server vs alternatives

Exa MCP Server and Tavily MCP focus on search for agents and return content from their own indexes, while Firecrawl MCP centers on scraping, crawling and interacting with pages you choose. Apify Actors MCP gives access to prebuilt scrapers for specific sites. For simple browser automation without a scraping API, Playwright MCP runs a local browser.

Troubleshooting Firecrawl MCP Server

• Only three tools appear: you are on the keyless hosted endpoint. Sign in through /v2/mcp-oauth or add an API key to unlock crawl, map, agent and the rest.
• Rate limit errors: the server surfaces Firecrawl rate-limit errors to the client. Signing in or using an API key gives higher limits; for crawls, also lower the limit.
• Windows issues with the Cursor command: try cmd /c "set FIRECRAWL_API_KEY=your-api-key && npx -y firecrawl-mcp".
• Opening the OAuth URL in a browser does nothing useful: it is a server URL; add it in your MCP client and use the client's account-connection flow.
• Too many tools for your client's tool limit: set FIRECRAWL_NO_SEARCH_FEEDBACK=1 and/or FIRECRAWL_NO_ENDPOINT_FEEDBACK=1, or use the search-only endpoint https://mcp.firecrawl.dev/v2/mcp-search.