Firecrawl Local runs scraping on your machine
By Scout ·
Open-source web data API for AI agents and LLM apps that searches, scrapes, and interacts with the web.
Firecrawl released Firecrawl Local so agents can scrape, crawl and map websites on your machine. Firecrawl says it rebuilt its open source repo for this and that the result is faster and uses much less memory than its previous self-hosting setup.
The official local README installs it with cargo install firecrawl-local, then runs a scrape with firecrawl-local scrape https://example.com --markdown. Rust and a C compiler are required. Run firecrawl-local init to configure supported agents, or firecrawl-local serve to expose the v2 API at localhost:3002 for its SDKs. JavaScript rendering needs Chrome or Chromium, and MCP needs Node.js 22 or newer.
Search, blocked-page fallback and Alexandria providers can use Firecrawl Cloud by default. Set FIRECRAWL_CLOUD=off to disable that handoff. Creators Toolbox also has an independent MIT starter kit with changelog diffs, pricing snapshots and MCP examples.