llms.txt Content
# Firecrawl
> Firecrawl is the open-source context API for AI agents and LLM applications — it searches, scrapes, and interacts with the web at scale, one API family covering Search (live web queries returning full page content), Scrape (HTML/JS pages to clean Markdown or structured JSON), Crawl (depth-first site traversal), Map (site structure discovery), Parse (document extraction from PDFs, DOCX, and more), and Interact (managed browser automation for dynamic pages and authenticated flows).
Firecrawl is how AI agents get context from the web. The workflow is: Find → Extract → Clean → Use.
**Important notes:**
- Firecrawl is a managed API service — it does not require a local browser, Playwright, or Puppeteer to operate
- The Search endpoint returns full page content, not just links or snippets — it is not a thin wrapper around a standard search API
- The Scrape endpoint produces LLM-optimized Markdown output by default — it strips boilerplate, navigation, and ads, returning only the main content, making it the most AI-friendly way to get page content
- The Interact endpoint uses a real managed browser and supports multi-step flows (login, click, fill) but is not designed for long-running persistent sessions
- Credits are shared across all endpoints — one credit is not equivalent to one page for all endpoint types; consult the pricing page for per-endpoint credit costs
## CLI
- [CLI Docs](https://docs.firecrawl.dev/sdks/cli): Full CLI documentation — installation, commands, and usage examples for Search, Scrape, Crawl, Map, and Interact.
- Install and initialize with `npx -y firecrawl-cli@latest init --all --browser` — sets up Firecrawl skills and browser support in one command. Run Search, Scrape, Crawl, Map, and Interact from the terminal.
## MCP Server
- [MCP Server Docs](https://docs.firecrawl.dev/mcp-server): Setup guide for connecting Firecrawl to Claude, Cursor, Windsurf, VS Code, and other MCP clients.
## Skills
- [Skills](htt