How to Set Up Firecrawl MCP in Cursor (Step by Step)
2026-08-07 · by bixzell
How to Set Up Firecrawl MCP in Cursor (Step by Step)
Firecrawl’s MCP server is hosted — there is no local server to install. You point Cursor at a URL and the scrape, search and interact tools show up. The keyless free tier even works without an API key.
7,148⭐ projects on GitHub — one of the most-used scraping servers in the ecosystem.
Prerequisites
- Cursor (any recent version with remote MCP support)
- No local install, no Node, no npm
Step 1 — Add the remote server
In Cursor, open Settings → MCP (or the command palette → cursor mcp), then add a remote / HTTP server:
- Name:
firecrawl - Type:
http(Streamable HTTP) - URL:
https://mcp.firecrawl.dev/v2/mcp
That’s the whole install. The server is served by Firecrawl, so it works from any machine.
Step 2 — Use the keyless free tier (or connect a key)
On the keyless tier, three tools work immediately, rate-limited:
scrape— turn any URL into clean markdownsearch— web search with full page contentinteract— operate pages (click, navigate)
Tools like crawl, map, agent and extract need an account. Two ways to connect:
- OAuth (recommended for interactive use): use the OAuth URL
https://mcp.firecrawl.dev/v2/mcp-oauthand sign in once. - API key: keep the base URL and set the header
Authorization: Bearer <FIRECRAWL_API_KEY>in the MCP server config.
There is also a read-only search-only endpoint at https://mcp.firecrawl.dev/v2/mcp-search if you only want search.
Step 3 — Verify
- Toggle the server on in MCP settings.
- It should show a green connected state.
- Test with: “Scrape https://example.com and summarize it.”
If tools appear and run, you’re done. For the full client matrix (Claude Desktop, VS Code, Codex, Windsurf), see the Firecrawl compatibility page.
Troubleshooting
| Symptom | Fix |
|---|---|
| Server not connecting | Double-check the URL is exactly https://mcp.firecrawl.dev/v2/mcp (no trailing slash issues) and the type is HTTP. |
crawl / map / agent require auth |
Those tools need an account — use the OAuth URL or add an API key header. |
| Rate limited on keyless tier | Sign up and add a key to raise limits. |
| Empty results on JS-heavy pages | Enable the headless browser crawl option in the Firecrawl dashboard — plain scrapes miss client-rendered content. |
Next steps
- If you’re choosing between Firecrawl and a search-style server, read Firecrawl vs Exa.
- For the ranked list of scraping alternatives, see Firecrawl MCP alternatives.