Firecrawl

Firecrawl is a web extraction API designed to feed AI systems: it transforms a page or entire website into clean markdown or JSON, handles JavaScript rendering, follows links to explore a domain, fills forms and clicks, and returns screenshots and metadata. It is the technical brick you place upstream of a search engine, knowledge base or agent where Browse AI targets the non-technical user. Billing is per credit: one page crawled costs one credit, a search costing two credits for ten results, one minute of interactive navigation costing two credits. The free plan includes 1,000 credits. Hobby costs 16 USD/month for 5,000 credits, Standard 83 USD/month for 100,000 credits, Growth 333 USD/month for 500,000 credits and Scale 599 USD/month for 1 million, amounts shown on annual billing.

Firecrawl is published open source and can be self-hosted, but the publisher honestly documents the limits: the self-hosted instance lacks the proprietary anti-blocking engine handling detection, some API endpoints do not work, and any collection method beyond basic modes must be hand-configured. The service has no French interface or docs. The publisher displays SOC 2 Type 2 certification and a non-retention option limited to the Enterprise plan, with no published EU hosting region. Official SDKs exist in Python, Node.js, Go, Rust, Java and Elixir, with a CLI tool and an MCP server usable from Claude, Cursor or Windsurf: the tool clearly targets developer profiles.