Firecrawl

Firecrawl turns websites into data that language models can use. Instead of writing and maintaining your own scrapers, you call one API and get clean Markdown, JSON or screenshots back.

Main endpoints

• Scrape: one URL to clean Markdown, structured JSON (via a schema or prompt), HTML or a screenshot. It handles JavaScript-rendered pages and also parses PDFs and DOCX files.
• Crawl and Map: follow links across a whole site, or list its URLs.
• Search: web search that returns the full content of results.
• Interact: click, scroll, type and wait on a page before extracting.
• Agent (preview): autonomous multi-step data gathering.

SDKs exist for Python and Node, along with a CLI and an MCP server, so you can plug Firecrawl into tools like Claude Code or Cursor.

Pricing

• Free plan: 1,000 credits a month, no card required (1 credit is roughly one scraped page).
• Paid plans: start with Hobby at $16/month billed yearly ($19 monthly).
• Open source: the core is AGPL-3.0 on GitHub and can be self-hosted; the cloud version has extra features.

Good for

Developers building RAG pipelines, research agents or anything that needs fresh web content in an LLM-friendly format.