Firecrawl MCP

What it does

• Scrape, crawl, map, and search any site through Firecrawl's API — output is clean, LLM-ready markdown
• Handles JS-rendered pages, sitemap discovery, and multi-page crawls with depth/limit controls
• Structured extraction modes turn pages into schema-shaped JSON, not just prose
• Official Firecrawl-maintained server

When to use it

Best when Claude needs real web pages as context — docs ingestion, competitor research, 'read this link first' flows on JS-heavy sites. Cheaper and steadier than a browser agent for read-only targets; switch to Playwright MCP or Browserbase when the page needs interaction.

Getting started

Add to .mcp.json under mcpServers.firecrawl:

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

Things to know

Firecrawl is a paid cloud service (free tier rate-limited) — page content transits their API, so keep private or authenticated pages out of scope. Deep crawls can trip target-site rate limits; set limits and mind robots.txt.