We just shipped a major upgrade to Firecrawl
search. Agents using it now achieve state-of-the-art accuracy on SimpleQA (94.7%) while using 10x fewer tokens than processing full pages.
Learn more
Built for AI agents that need highly relevant context
/search now runs every result through a custom relevance model, scoring each paragraph, list, and table against your query. Agents only receive the excerpts that best answer it, with no tokens wasted on noisy pages.
Our improved /search is on by default for free, and is live for every Firecrawl user across the API, SDKs, CLI, and MCP.