Firecrawl

Search, scrape, map, crawl and interact endpoints

The API covers web search with full result content, scraping pages into Markdown, JSON, or screenshots, mapping and crawling sites, and an Interact feature that lets a scraped page be operated via AI prompts or code.

Handles JS rendering, PDFs and page actions

Zero-configuration handling of JavaScript rendering, smart waiting for content to load, and parsing of PDFs, DOCX and other media. Actions support clicking, scrolling, typing, waiting, key presses and screenshots on live pages.

Published benchmarks and token savings

The site reports 96% coverage and P95 latency of 3,387 ms on a 1,000-URL benchmark run in January 2026, plus roughly 93% fewer input tokens by stripping navigation, footers and ads from output.

SDKs, MCP and open source repo

Python, Node.js, cURL and CLI usage are shown, and agents can connect through a CLI init command or MCP client. The project is open source with a public GitHub repository and contributor community.