| name | mcp-firecrawl |
| description | This skill provides guidance for using Firecrawl MCP for web scraping and content extraction. Use when scraping web pages, crawling websites, searching the web with content extraction, mapping site URLs, or extracting structured data from web pages. |
Firecrawl MCP
Overview
This skill enables web scraping and content extraction using Firecrawl through MCP. Firecrawl provides powerful capabilities for scraping single pages, crawling entire websites, searching with content extraction, and extracting structured data.
When to Use This Skill
- Scraping content from single web pages
- Crawling multiple pages on a website
- Searching the web with automatic content extraction
- Discovering all URLs on a website
- Extracting structured data (prices, names, details)
Tool Selection Guide
| Task | Tool | Notes |
|---|
| Single page content | firecrawl_scrape | Fast, reliable |
| Find URLs on site | firecrawl_map | Discovery only |
| Web search | firecrawl_search | With optional scraping |
| Multiple pages | firecrawl_map + firecrawl_scrape | Better than crawl |
| Structured data | firecrawl_extract | Uses LLM extraction |
| Full site crawl | firecrawl_crawl | Use with caution |
Core Tools
firecrawl_scrape
Scrape content from a single URL. Most powerful and reliable scraper.
Tool: mcp__firecrawl__firecrawl_scrape
mcp__firecrawl__firecrawl_scrape({
url: "https://docs.example.com/api",
formats: ["markdown"],
onlyMainContent: true,
maxAge: 172800000
})
Key Parameters:
url: Target URLformats: Output formats - "markdown", "html", "links"onlyMainContent: Extract only main content (recommended)maxAge: Cache TTL in ms (500% faster with cache)
firecrawl_search
Search the web and optionally scrape results.
Tool: mcp__firecrawl__firecrawl_search
Without scraping (preferred):
mcp__firecrawl__firecrawl_search({
query: "xterm.js terminal tutorial",
limit: 5
})
With scraping:
mcp__firecrawl__firecrawl_search({
query: "VS Code extension API",
limit: 3,
scrapeOptions: {
formats: ["markdown"],
onlyMainContent: true
}
})
Search Operators:
| Operator | Example | Description |
|---|
"" | "exact phrase" | Exact match |
- | -deprecated | Exclude term |
site: | site:github.com | Specific site |
inurl: | inurl:docs | URL contains |
intitle: | intitle:tutorial | Title contains |
firecrawl_map
Discover all URLs on a website.
Tool: mcp__firecrawl__firecrawl_map
mcp__firecrawl__firecrawl_map({
url: "https://docs.example.com",
limit: 100,
search: "api"
})
firecrawl_extract
Extract structured data using LLM.
Tool: mcp__firecrawl__firecrawl_extract
mcp__firecrawl__firecrawl_extract({
urls: ["https://shop.example.com/product/123"],
prompt: "Extract product details",
schema: {
type: "object",
properties: {
name: { type: "string" },
price: { type: "number" }
},
required: ["name", "price"]
}
})
firecrawl_crawl
Crawl multiple pages (use sparingly).
Tool: mcp__firecrawl__firecrawl_crawl
mcp__firecrawl__firecrawl_crawl({
url: "https://docs.example.com/guide",
maxDiscoveryDepth: 2,
limit: 10
})
Warning: Crawl can return large responses. Keep limits low.
Best Practices
- Prefer scrape over crawl: Use
map + scrape for better control
- Use caching: Add
maxAge for repeated requests
- Search then scrape: Search without formats, then scrape relevant URLs
- Limit crawl depth: Keep
maxDiscoveryDepth and limit low
- Use onlyMainContent: Get cleaner content without navigation
References
For complete tool parameters, see references/tools.md.
