Firecrawl Synced

Connect Claude Code to Firecrawl for intelligent web scraping, content extraction, and structured data retrieval from websites through the Model Context Protocol.

When to Use This MCP Server

Connect this server when you need to:

• Scrape web pages and extract clean, structured content from websites for data analysis, content migration, or reference during development
• Crawl entire websites or sitemaps to build comprehensive content inventories, verify link structures, or gather competitive intelligence
• Convert web pages into clean markdown or structured data formats for ingestion into documentation systems or knowledge bases

Consider alternatives when:

• You need simple single-page fetching without JavaScript rendering, where a basic HTTP request or the built-in WebFetch tool suffices
• Your scraping target requires browser automation with complex interaction sequences like login flows, where Playwright or Puppeteer is more appropriate

Quick Start

Configuration

Copy
name: firecrawl-synced
type: mcp
category: devtools

Example Connection

Copy
claude mcp:connect firecrawl-synced

Available Tools

scrape:            Scrape a single URL and return clean content in markdown or structured format
crawl:             Crawl multiple pages starting from a URL with depth and scope controls
search:            Search the web and return scraped content from top results
extract:           Extract structured data from pages using natural language schemas
map:               Discover all URLs on a website and return a sitemap-style listing

Core Concepts

Firecrawl MCP Server Overview

Aspect	Details
Transport	stdio-based process via npx running firecrawl-mcp package
Authentication	Firecrawl API key for accessing the hosted scraping infrastructure
Rendering	Full JavaScript rendering via headless browsers for dynamic content
Output Formats	Clean markdown, HTML, structured JSON, and screenshot captures
Rate Handling	Automatic rate limiting, retry logic, and concurrent request management

Firecrawl Architecture

Claude Code IDE
    |
    v
[MCP Client] --stdio--> [firecrawl-mcp]
                              |
                              v
                     [Firecrawl Cloud API]
                              |
                    +---------+---------+
                    |         |         |
                    v         v         v
              [Scrape    [Crawl     [Extract
              Engine]    Engine]    Engine]
                    \         |         /
                     [Headless Browser Pool]
                      JS Render | Anti-Bot

Configuration

Parameter	Type	Default	Description
FIRECRAWL_API_KEY	string	required	Firecrawl API key from your Firecrawl dashboard
default_format	string	markdown	Default output format: markdown, html, or json
max_pages	integer	50	Maximum pages to crawl per crawl operation
wait_for_js	boolean	true	Wait for JavaScript execution before extracting content
timeout	integer	30000	Timeout in milliseconds for individual page scraping operations

Best Practices

1. Use Scrape for Single Pages, Crawl for Sites - The scrape tool is optimized for individual page content extraction while crawl handles multi-page traversal. Using crawl for a single page adds unnecessary overhead; use scrape instead.

2. Define Extraction Schemas for Structured Data - When using the extract tool, provide clear natural language descriptions of the data structure you need. Well-defined schemas produce more accurate and consistent structured data output.

3. Set Crawl Depth and Scope Limits - Always configure max_pages and depth limits when crawling. Without limits, a crawl operation can traverse thousands of pages, consuming API credits and taking significant time to complete.

4. Prefer Markdown Output for Content Analysis - Markdown output strips navigation, ads, and boilerplate content, giving you clean readable text. Use HTML output only when you need to preserve the exact page structure and styling.

5. Handle Rate Limits Gracefully - Firecrawl manages rate limiting automatically, but large crawl operations may take time. Monitor crawl progress and consider breaking large sites into targeted section crawls for faster results.

Common Issues

1. Scrape Returns Empty Content - The target page may use aggressive anti-bot measures or require authentication. Check if the page loads content dynamically and ensure wait_for_js is enabled for JavaScript-heavy sites.

2. Crawl Exceeds API Credit Limits - Large crawl operations consume credits proportional to pages crawled. Set conservative max_pages limits and use the map tool first to understand site size before initiating a full crawl.

3. Extracted Data Missing Expected Fields - Refine your extraction schema with more specific field descriptions. The extract tool relies on natural language understanding of your schema, so ambiguous field names produce inconsistent results.