name: firecrawl description: Uses Firecrawl to scrape web pages to clean markdown, search and scrape top results, crawl entire websites, or map a domain. Use when the user needs to scrape a URL, crawl a site, search the web and get page content, or discover/map URLs on a domain.

Firecrawl

When to Use

• Scrape a single page to clean markdown for LLMs or processing
• Search the web and scrape the top results (query → markdown)
• Crawl an entire website with limits and timeout
• Map a domain to discover/index URLs (search, sitemap options)

Setup

API key: set FIRECRAWL_API_KEY in .env (or environment). Get a key at firecrawl.dev.

Project helper (recommended): use firecrawl_tools.scrape_url, search_and_scrape, crawl_site, map_domain — they read the key from env and return errors if unset.

Direct SDK (firecrawl-py v4):

import os
from firecrawl import FirecrawlApp

app = FirecrawlApp(api_key=os.getenv("FIRECRAWL_API_KEY"))

Scrape One URL (v4: scrape)

Returns a Document (markdown, metadata).

doc = app.scrape("https://example.com", only_main_content=True)
# doc has markdown/content and metadata

Or use project helper: firecrawl_tools.scrape_url(url).

Search and Scrape Top Results (v4: search)

result = app.search("what is Firecrawl?", limit=5)
# result is SearchData (scraped content for top results)

Or: firecrawl_tools.search_and_scrape(query, limit=5).

Crawl a Website (v4: crawl)

Starts crawl and waits until done or timeout. Returns CrawlJob (status, data).

job = app.crawl("https://example.com", limit=100, timeout=300)
# job.status, job.data

Or: firecrawl_tools.crawl_site(start_url, limit=100, timeout=300).

To start without waiting, use app.start_crawl(url, limit=...) then app.get_crawl_status(job_id) to poll.

Map a Domain (v4: map)

Discover URLs on a domain (optional search query, sitemap, limit).

map_result = app.map("https://example.com", search="pricing", limit=50)

Or: firecrawl_tools.map_domain(url, search=..., limit=...).

CLI (Optional)

User can run locally:

npx -y firecrawl-cli@latest init --all --browser

After that, the CLI can scrape/crawl from the command line; the agent can suggest CLI commands when appropriate.

Notes

• Prefer reading FIRECRAWL_API_KEY from environment; do not hardcode keys.
• For LLM extraction with a schema, use app.extract (v4) or see Firecrawl docs.