Overview

The Firecrawl MCP Server integrates Firecrawl's powerful web data platform with AI assistants. It enables searching, scraping, crawling, and extracting structured data from websites — all optimized for consumption by LLMs.

Official Server:

Developed and maintained by Firecrawl

Key Features

🔍

Web Search

Search the web and optionally extract content from search results

📄

URL Scraping

Extract clean markdown or structured JSON from any URL

🕸️

Site Crawling

Crawl entire websites with depth and page limits

🗺️

URL Discovery

Map websites to discover all indexed URLs before scraping

🤖

Autonomous Agent

AI agent that independently browses and gathers data from multiple sources

📊

Structured Extraction

Extract specific data using JSON schemas and LLM-powered parsing

Available Tools

Quick Reference

Tool	Purpose	Category
`firecrawl_search`	Search the web for info	Search
`firecrawl_scrape`	Extract content from a single URL	Content
`firecrawl_crawl`	Multi-page extraction with limits	Crawl
`firecrawl_map`	Discover URLs on a site	Discovery
`firecrawl_agent`	Complex multi-source research	Agent
`firecrawl_extract`	Structured data extraction with schema	Extract

Detailed Usage

firecrawl_scrape▶

Extract clean content from a single URL in markdown or structured JSON format.

{
  "url": "https://example.com/product",
  "formats": ["markdown"],
  "onlyMainContent": true
}

firecrawl_search▶

Search the web and optionally extract content from the top results.

{
  "query": "latest AI research papers 2025",
  "limit": 5,
  "lang": "en",
  "scrapeOptions": {
    "formats": ["markdown"],
    "onlyMainContent": true
  }
}

firecrawl_crawl▶

Crawl a website with configurable depth and page limits. Returns a job ID for status polling.

{
  "url": "https://example.com/docs/*",
  "maxDepth": 2,
  "limit": 50
}

firecrawl_agent▶

Autonomous research agent that browses the web independently. Returns a job ID; poll with firecrawl_agent_status.

{
  "prompt": "Find the top 5 AI startups founded in 2025 and their funding amounts",
  "schema": {
    "type": "object",
    "properties": {
      "startups": {
        "type": "array",
        "items": {
          "type": "object",
          "properties": {
            "name": { "type": "string" },
            "funding": { "type": "string" }
          }
        }
      }
    }
  }
}

Installation

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

API Key Required:

Get your API key from firecrawl.dev

Tool Selection Guide

Know the exact URL? Use scrape (JSON format for structured data)
Need to discover URLs? Use map first
Multiple known URLs? Use batch_scrape
Open-ended question? Use search
Complex multi-source research? Use agent
Entire site analysis? Use crawl (with limits)

Sources

Blender MCP Server

Blender MCP Server connects Blender to Claude AI through the Model Context Protocol (MCP), enabling prompt-assisted 3D modeling, scene creation, and manipulation.

Context7 MCP Server

Context7 MCP server provides up-to-date, version-specific code documentation directly in LLM context, preventing hallucinated APIs and outdated code generation.

Airtable MCP Server

Airtable MCP servers enable AI models to interact with Airtable bases, providing capabilities for base management, record operations, field configuration, and structured data automation.

Firecrawl MCP Server