research Official server available

Firecrawl MCP Server

The Firecrawl MCP lets Claude scrape, search and crawl any site. How marketers use it for competitor and content research, with hosted and local setup.

Pricing
freemium
GitHub stars
7,500
Last commit
2026-09-22
Page updated

Free plan includes 1,000 credits a month (1 credit per scraped page); paid plans start at $16/month billed annually for 5,000 credits.

Firecrawl's MCP server is how I get Claude to actually read the web instead of guessing. Give it a URL and firecrawl_scrape returns clean markdown, or JSON that matches a schema you describe. firecrawl_map lists every URL on a site, firecrawl_crawl pulls a whole section, and firecrawl_search finds pages and scrapes them in one call.

Firecrawl builds this server itself and hosts it at https://mcp.firecrawl.dev/v2/mcp. You can try it with no API key at all: scrape, search and parse work on a rate-limited keyless tier. Add a key (or sign in with OAuth) and you get the full set, including crawl, map, the research agent and page monitors. There's also a search-only version in Claude's connector directory if you just want safe, read-only search.

My go-to competitor workflow: map a competitor's site, filter to /pricing, /features and /customers, scrape those pages as JSON with fields like plan name, price and limits, and have Claude build a comparison table. Then set a firecrawl_monitor on their pricing page so you hear about changes. It pairs well with the Perplexity MCP for the "why" behind the numbers.

Caveats: crawls burn credits fast, so always set a limit. And scraping is on you. Respect robots.txt and each site's terms, especially for anything you publish.

Official vs community Firecrawl MCP

Firecrawl hosted MCP

Remote server URL: https://mcp.firecrawl.dev/v2/mcp

The GitHub server is Firecrawl's own, and the hosted URL runs the same code, so there's no community fork worth choosing over it. Use the hosted URL if your client can send an Authorization header or do OAuth; run npx firecrawl-mcp locally for Claude Desktop or if you self-host Firecrawl via FIRECRAWL_API_URL.

What you can ask it

  • “Map competitor.com, scrape their pricing and features pages as JSON (plan, price, seats, limits), and build a comparison table against ours”
  • “Crawl the /blog section of competitor.com, limit 50 pages, and list their topics grouped by funnel stage with publish dates”
  • “Search for "best email marketing tools 2026", scrape the top 10 results, and tell me which tools every list includes”
  • “Scrape these 5 landing pages and pull the headline, subhead, primary CTA and social proof from each”
  • “Set a monitor on competitor.com/pricing and tell me what changed since last week”

Features

  • `firecrawl_scrape`: any URL to clean markdown, or JSON matching a schema you supply
  • `firecrawl_map`: list every URL on a site without fetching content
  • `firecrawl_crawl`: pull many pages under a site with path filters and page limits
  • `firecrawl_search`: web search with optional full-page content in the same call
  • `firecrawl_agent`: multi-source research that returns structured data when you don't know the URLs
  • `firecrawl_interact`: click, type and move through pages that hide content behind actions
  • `firecrawl_monitor_*`: scheduled checks on a page with diffs and change alerts

How to set up the Firecrawl MCP Server

Works with: Claude, Claude Code, Claude Desktop, Cursor, VS Code, Windsurf

1. Get an API key at firecrawl.dev/app/api-keys. (Optional: skip this and use the keyless hosted URL for scrape, search and parse only.) 2. Claude Code, hosted server:
claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp --header "Authorization: Bearer fc-YOUR_API_KEY"
3. Claude Code, local server:
claude mcp add firecrawl -e FIRECRAWL_API_KEY=fc-YOUR_API_KEY -- npx -y firecrawl-mcp
4. Claude Desktop, Cursor, Windsurf: add this to your MCP config:
{
  "mcpServers": {
    "firecrawl": {
      "command": "npx",
      "args": [
        "-y",
        "firecrawl-mcp"
      ],
      "env": {
        "FIRECRAWL_API_KEY": "YOUR_API_KEY_HERE"
      }
    }
  }
}
5. claude.ai: add Firecrawl from Claude's connector directory. That connector is the read-only search endpoint (/v2/mcp-search); for scraping and crawling use one of the setups above. 6. Never paste your API key into the server URL or a chat. Keep it in the header or env var.

Alternatives

Playwright MCP

Free local browser automation from Microsoft; more setup and no clean markdown or JSON extraction built in.

Brave Search MCP

Cheaper if you only need search results and snippets, not full page content.

FAQ

Is there an official Firecrawl MCP server?
Yes. firecrawl/firecrawl-mcp-server is maintained by Firecrawl, published on npm as firecrawl-mcp, and hosted at https://mcp.firecrawl.dev/v2/mcp.
How do I connect Firecrawl to Claude?
In Claude Code, run claude mcp add --transport http firecrawl https://mcp.firecrawl.dev/v2/mcp --header "Authorization: Bearer fc-YOUR_API_KEY". In Claude Desktop, add npx -y firecrawl-mcp with FIRECRAWL_API_KEY to your config.
Is the Firecrawl MCP free?
The server is open source, and the hosted keyless tier lets you scrape and search without an account at a low rate limit. With an account you get 1,000 free credits a month, one credit per page scraped.
Can I use Firecrawl for competitor research?
Yes, that's one of the best uses: map a competitor's site, scrape pricing and feature pages into structured JSON, and set monitors to catch changes. Stick to public pages and respect each site's terms.

Added: 2026-09-22 · Last reviewed: 2026-09-22