Source: Firecrawl docs (introduction and scrape)
Firecrawl's docs for turning websites into LLM-ready markdown or JSON: scrape, crawl, map, search, interact and agent endpoints, caching via maxAge and credit costs.
Contents
- What: official documentation (two pages)
- Author / publisher: Firecrawl
- Date: read 2026-11-14 (living pages)
- Original: Introduction, Scrape
Summary
Firecrawl is a hosted (and self-hostable open-source) scraping service aimed at LLM apps. It handles JavaScript rendering, proxies and anti-bot measures and returns clean markdown, HTML, screenshots or schema-shaped JSON.
Key claims
- Endpoints: Scrape (one URL), Crawl (recursive site), Map (discover URLs), Search (web search returning full page content), Interact (click, type, fill forms on a scraped page), Agent (autonomous data gathering), Parse (local PDF/DOCX/XLSX to markdown).
- Formats: markdown, summary, HTML, raw HTML, links, images, screenshot (URLs expire after 24 h), JSON from a schema or prompt, and others.
- Actions (click, type, wait, screenshot) can run before scraping; Interact is recommended for richer control.
- Caching:
maxAgedefaults to 2 days and can make scrapes up to 5x faster;maxAge: 0forces fresh fetches (slower, more failures);minAgegives cache-only lookups; change tracking bypasses the cache. - PDFs and DOCX at URLs are auto-detected and converted to markdown.
- Costs: 1 credit per scrape (cached or not); JSON extraction +4; PDF 1 credit per page; zero data retention +1 per page.