firecrawl/firecrawl-mcp-server

Hosted: https://mcp.firecrawl.dev/v2/mcp (keyless free tier; OAuth at /v2/mcp-oauth, API key via Authorization: Bearer header)
npx: env FIRECRAWL_API_KEY=fc-YOUR_API_KEY npx -y firecrawl-mcp
npm install -g firecrawl-mcp
Streamable HTTP local mode: env HTTP_STREAMABLE_SERVER=true FIRECRAWL_API_KEY=... npx -y firecrawl-mcp (URL http://localhost:3000/mcp)

Usage

• •Configure the server in Cursor (v0.45.6+), Windsurf, VS Code, or Claude Desktop with FIRECRAWL_API_KEY set
• •Choose the right tool: scrape for a known URL, map to discover URLs, search for open-ended queries, crawl for multi-page extraction, agent for multi-source research, interact for browser automation
• •Prefer JSON format in firecrawl_scrape with a schema (avoids context-window overflow); use markdown only when full content is needed
• •Call firecrawl_search_feedback after using search results to refund 1 credit and improve search quality

Key features

• ✓Tools: firecrawl_scrape, firecrawl_map, firecrawl_search, firecrawl_crawl, firecrawl_parse, firecrawl_extract, firecrawl_agent (async, polled), firecrawl_interact (browser automation), firecrawl_monitor_* (recurring page monitors), firecrawl_research_* (papers/GitHub), firecrawl_developer_search (GitHub issues/PRs/READMEs/docs index)
• ✓Developer search index built for coding agents: GitHub issues, merged PRs, READMEs, curated docs
• ✓Keyless free tier lets scrape/search/interact work without an API key (rate-limited); crawl, map, agent, extract still need a key
• ✓Automatic retries and rate limiting; self-hosted support via FIRECRAWL_API_URL; SSE and Streamable HTTP transports
• ✓Search feedback tools refund credits and improve search quality

Best for

Giving coding agents a reliable web-scraping/search layer with clean structured output, including a developer-focused index for programming questions.

Caveats

• ⚠Keyless free tier is rate-limited; full tool set and higher limits require a Firecrawl API key or OAuth login
• ⚠firecrawl_crawl responses can be very large and exceed token limits - limit depth and page count
• ⚠Hosted firecrawl_parse cannot read the caller's filesystem directly (two-step upload-ref flow); local file parsing requires a self-hosted FIRECRAWL_API_URL
• ⚠Never put an API key in the server URL; use headers or env vars