Official · by firecrawl

Firecrawl MCP server

The Firecrawl MCP server lets an AI assistant search, scrape, and interact with web content, plus search papers.

  • Hosted (remote) · Runs locally
  • MIT licence
  • ★ 7,635 on GitHub
  • Verified against its source 19 Oct 2026

What the Firecrawl MCP server does

An assistant can scrape a single URL, map a site to list its indexed URLs, search the web, and start crawl jobs that return final status and data. It can also parse local files or hosted upload references, run an autonomous research agent when URLs are unknown, and interact with pages. A remote endpoint is listed at https://mcp.firecrawl.dev/v2/mcp, and a local option runs via npx firecrawl-mcp with a Firecrawl API key.

Tools it gives your assistant

10 tools or toolsets, as its documentation lists them.

  • firecrawl_scrape Scrapes content from one URL with options
  • firecrawl_map Lists the URLs indexed on a website
  • firecrawl_search Searches the web and can extract page content from results
  • firecrawl_crawl Crawls many related pages and returns the final status and data
  • firecrawl_parse Parses local or uploaded documents into markdown or JSON
  • firecrawl_agent Runs autonomous research that returns structured data across sites
  • firecrawl_interact Clicks, types and navigates dynamic pages in a browser
  • firecrawl_interact_stop Ends an interact session for a scraped page
  • firecrawl_gov_search Searches US government legal and regulatory sources
  • firecrawl_research_*toolset Searches and reads scientific papers and repositories

How to install it

Hosted server

Add this address to your assistant's MCP settings:

https://mcp.firecrawl.dev/v2/mcp

Run it locally

With Node.js (npx)

npx -y firecrawl-mcp

Command

npm install -g firecrawl-mcp

What you need

In the firecrawl documentation's own words:

  • “Replace your-api-key with your Firecrawl API key”

Commands and settings change between versions. Check the official instructions before you install.