Firecrawl MCP: Web Content Extraction

What Firecrawl does

Firecrawl MCP converts web pages into clean markdown. Give it a URL and it returns the page content without navigation bars, cookie banners, sidebars, or advertising. This is useful when the agent needs to read a specific documentation page, analyze a competitor's API, or extract structured data from a web resource.

Firecrawl handles JavaScript-rendered pages, so it works with single-page applications and dynamically loaded content that basic HTTP fetches would miss.

Setup

Firecrawl MCP uses the stdio transport. Get an API key from firecrawl.dev (free tier available).

# Add Firecrawl MCP
autohand mcp add firecrawl npx -y firecrawl-mcp

# Set your API key
export FIRECRAWL_API_KEY="your-firecrawl-api-key"

# Verify the connection
autohand mcp list

Or configure it in ~/.autohand/config.json:

{
  "mcp": {
    "servers": [
      {
        "name": "firecrawl",
        "transport": "stdio",
        "command": "npx",
        "args": ["-y", "firecrawl-mcp"],
        "env": {
          "FIRECRAWL_API_KEY": "your-firecrawl-api-key"
        },
        "autoConnect": true
      }
    ]
  }
}

Usage examples

Read a documentation page

# Point the agent at a specific doc page
autohand --prompt "Read https://docs.stripe.com/payments/accept-a-payment and implement a payment form using their latest API"

# Firecrawl extracts the Stripe docs as markdown, the agent uses it to write code

Extract API reference details

# Pull in a full API reference
autohand --prompt "Read the Cloudflare Workers API docs at https://developers.cloudflare.com/workers/runtime-apis/ and create a KV-backed REST API"

# Gets the full reference, not just what the model remembers from training

Analyze a GitHub README

# Read a project's README for context
autohand --prompt "Read https://github.com/drizzle-team/drizzle-orm and set up Drizzle with our existing PostgreSQL database"

Crawl multiple pages

# Firecrawl can crawl linked pages from a starting URL
autohand --prompt "Crawl the Hono.js documentation starting from https://hono.dev/docs/ and build a REST API with authentication middleware"

Firecrawl vs basic fetch

Autohand has a built-in fetch_url tool. Here is when to use Firecrawl instead:

Use the built-in fetch for simple static pages. Use Firecrawl when you need clean output from complex sites or need to crawl multiple linked pages.

Tips

• Firecrawl's free tier gives you 500 pages per month. For heavy documentation crawling, consider their paid plans.
• When crawling, set a depth limit to avoid pulling in hundreds of pages. The agent is smart enough to stop when it has enough context.
• Pair Firecrawl with Brave Search MCP: search to find the right page, then Firecrawl to read it in full.
• Firecrawl respects robots.txt and handles rate limiting automatically.