releases.sh
Firecrawl

Firecrawl

firecrawl.devDeveloper Tools
$npx @buildinternet/releases get firecrawl
Recently Shipped8 releases · updated Aug 22, 2026

Developer Index launched for coding agents, covering 70M+ artifacts with the highest recall of any coding-specific index.

Developer Index targets coding agents. 70M+ artifacts span READMEs, docs, issues, PRs, and OpenAPI specs, refreshed continuously. It scores 0.63 recall@10 on the open DevDex benchmark, ahead of every other provider tested. Results carry stable ids, URLs, and matched passages in markdown, so agents answer directly without a second scrape.

Research Index added Life Sciences and went free. Search 41M+ papers across drug discovery, clinical trials, and biology, refreshed daily, with 90% recall@10. The existing AI/ML coverage — 3M+ arXiv papers and GitHub artifacts (issues, merged PRs, READMEs) — stays at 53.3% recall on arXivQA, beating the next best provider by 18%. Agents search abstracts, pull full text on demand, and verify claims against the source, and every category is now free.

Most accurate /search yet. A custom model scores every paragraph, list, and table against the query, returning the most relevant excerpts. Agents using it score 94.7% on SimpleQA — higher than any other provider — while using 10x fewer tokens than processing full pages. The API shape is unchanged, so existing calls improve automatically.

/monitor watches the web or single pages. The web-scale endpoint takes search queries and a plain-English goal, alerting via webhook or email when new results match, handling search, deduplication, filtering, and alerts in one call. The URL-based /monitor notifies on page changes, with up to 90% fewer tokens by ingesting only diffs, plus permalinks for every change.

Scraping improvements across formats. highlights returns only matching sentences, code blocks, and table rows (up to 100x fewer tokens); deterministicJson returns structured JSON without an LLM per request; lockdown: true serves results exclusively from Firecrawl's index; and the video format discovers videos on any page. /parse handles PDFs, Word docs, and spreadsheets up to 50 MB via a Rust engine up to 5x faster. Keyless access now covers /scrape, /search, /interact, and /parse from official MCP, CLI, and SDK clients.

AI-generated summaries may contain mistakes.

Sources

Latest releases

See all releases

Activity

122 releases14d interval2/mo
Mon
Wed
Fri
SepOctNovDecJanFebMarAprMayJunJulAug
Less
More