Firecrawl is a web data API for AI agents. An API (application programming interface) is an interface that another program calls over the internet. Firecrawl takes a web address or a search query and returns the page content in a form that a large language model can read directly: clean Markdown, structured JSON, HTML or screenshots.
Its main endpoints are Search (search the web and return the full content of the results), Scrape (convert one URL), Crawl (scrape all URLs of a website in one request) and Map (list the URLs of a site). It renders JavaScript, waits for content to load, can click, scroll and type before extracting, and parses PDF and DOCX files. SDKs exist for Python and Node.js, and AI agents connect through MCP (Model Context Protocol) or a command-line tool.
The code is open source, mainly under the AGPL-3.0 license with the SDKs under MIT, and the repository has a self-hosting guide. Firecrawl also runs a hosted cloud version with extra features, sold as monthly credit subscriptions with a small free plan.