fastCRW

A web data API built for AI agents and the teams shipping them. It turns searches and URLs into clean markdown, JSON, or HTML, so your agent gets usable content instead of raw pages. It's written in Rust and licensed AGPL-3.0, and you can use the managed cloud or run it yourself.

• Search: multi-engine web search that returns full page content.
• Scrape: full JS rendering, with actions like click, scroll and wait before extraction.
• Crawl and map: pull a whole site as structured data, or list every URL on a domain.
• Extract: get structured JSON from a prompt or a JSON schema.
• Monitor: watch a page and get an AI verdict plus a webhook when it changes.
• Parse: turn PDFs into markdown on CPU, with no LLM involved.

The cloud side handles proxies, headless browsers, anti-detection and scaling, so there's nothing to babysit. Every response includes source URLs and metadata, which makes answers easy to verify and debug.

It works with Python, Node, cURL and MCP, and plugs into Claude Code, Cursor, LangChain, n8n and similar tools. It's compatible with Firecrawl on the shared endpoints, though response fields differ slightly, so migration usually needs small adjustments. If you're comparing options, it's also a fit as a Tavily alternative.

Pricing is usage-based. The free tier gives 500 credits, with up to 500 more to earn, and no card. Paid plans start at $13/mo. One credit covers one page, search costs two, and failed requests aren't billed. The public benchmark harness shows higher recall than Firecrawl on the ArXivQA research set.