Firecrawl Changelog
Firecrawl shifted toward agent-driven automation and multi-step workflows. The /interact endpoint turned scrapes into persistent browser sessions where agents can click, fill forms, and navigate using natural language or code, with sessions reusable across calls. Parallel agent execution scaled data extraction from hours to minutes by running thousands of queries simultaneously with intelligent fallback to deeper research only when needed. The new Rust-based PDF parser added three configurable modes—fast, auto (with fallback OCR), and OCR-only—to handle everything from clean text to scanned documents, and new SDKs for Java and Elixir joined the lineup alongside a Firecrawl Skill that lets Claude, Antigravity, and other agents access scrape and search endpoints directly.
Firecrawl shipped agent-driven browser automation with the /interact endpoint, letting developers describe actions in natural language and have agents execute them across live sessions with persistent login state. The month also expanded language support with a full Java SDK covering core scraping and crawling operations.