Skip to main content

Firecrawl MCP Server Setup: Claude, Cursor, VS Code 2026

· 4 min read
Yassine El Haddad
Software Developer & Automation Specialist

I build production AI agents, web scrapers, and automation pipelines. Most of what I publish here comes from the actual problems they run into: proxies that get banned, anti-bot stacks that fingerprint your client, RAG that drifts when the underlying data moves. Stack: Python, TypeScript, Go, FastAPI, LangChain, Crawlee, Playwright, deployed on AWS, GCP, and Cloudflare.

The Firecrawl MCP server gives AI agents (Claude, Cursor, Windsurf, VS Code) web scrape, search, crawl, and extraction tools over the Model Context Protocol. Install in under two minutes with an API key and a JSON config block.

Get your Firecrawl API key →

What MCP Adds

Without MCP, Claude and Cursor cannot directly fetch web content. With Firecrawl MCP, agents can:

  • Scrape URLs to markdown
  • Search the web and scrape results
  • Map domains for URL discovery
  • Extract structured data with schemas
  • Run batch scrapes with retries

Prerequisites

  • Firecrawl account and API key
  • Node.js (for npx)
  • Claude Desktop, Cursor 0.48.6+, or VS Code with MCP support

Cursor Setup (0.48.6+)

  1. Open Cursor Settings → Features → MCP Servers
  2. Add this config:
{
"mcpServers": {
"firecrawl-mcp": {
"command": "npx",
"args": ["-y", "firecrawl-mcp"],
"env": {
"FIRECRAWL_API_KEY": "fc-YOUR-API-KEY"
}
}
}
}
  1. Restart Cursor

Alternatively, install the Firecrawl Cursor plugin from the marketplace for one-click setup.

Claude Desktop Setup

  1. Locate your Claude config (e.g. ~/Library/Application Support/Claude/claude_desktop_config.json on macOS)
  2. Add the same mcpServers block
  3. Store the API key in env or as a system environment variable
  4. Restart Claude Desktop

Available MCP Tools

Once connected, agents can use tools such as:

  • scrape — Single-URL scrape to markdown/HTML
  • crawl — Multi-page crawl with limits
  • map — Discover URLs on a domain
  • search — Web search with optional scrape
  • extract — Structured extraction with JSON schema
  • batch_scrape — Multiple URLs with retries

Exact tool names may vary by MCP package version; check the tool list in your client.

Example Prompts

  • "Scrape https://docs.firecrawl.dev and summarize the main features"
  • "Search for Firecrawl pricing 2026 and extract the plan names"
  • "Map all URLs on example.com that contain 'blog'"

Security and Boundaries

  • Use read-only tools first
  • Log tool calls (timestamp, source prompt) for audit
  • Keep input constraints explicit
  • Avoid mixing dev and prod API keys

Cost Awareness

MCP tool calls consume Firecrawl credits like direct API calls. Monitor usage in the Firecrawl dashboard. Free tier: 500 credits, 2 concurrent.

Apify Affiliate Banner 728x90Apify Affiliate Banner 728x90Apify Affiliate Banner 300x50Apify Affiliate Banner 300x50
Next step

Combine Firecrawl MCP with Make.com automation for scheduled scraping pipelines.

Frequently Asked Questions

Yes. Use npx -y firecrawl-mcp or the Cursor marketplace plugin. Pass FIRECRAWL_API_KEY in env.

Start with read-only tools, strict input validation, and full call logging before adding broader capabilities.

Check config path, env vars, and restart. Most issues are config or key-loading related.

Yes. Cursor 0.48.6+ supports MCP. Add the firecrawl-mcp block to Features > MCP Servers or use the marketplace plugin.

Common mistakes and fixes

Tools not appearing in Claude or Cursor

Restart the client after config. Verify MCP config path and FIRECRAWL_API_KEY in env. Cursor 0.48.6+ uses Features > MCP Servers.

MCP connection failed

Ensure npx is available. Try npx -y firecrawl-mcp manually. Check firewall for local MCP ports.