What it does
Firecrawl turns a URL (or a domain, or a search query) into LLM-ready Markdown, JSON, HTML, screenshots, or structured data via a schema. The API handles JavaScript rendering on the way in and Markdown conversion on the way out, so the model on the other end gets text it can ingest without an HTML-to-text pass. Six surfaces: Scrape, Crawl, Search, Map, Extract, and Actions (click/scroll/type for pages that need an interaction first). The pitch on the firecrawl.dev homepage is "Power AI agents with clean web data" and the customer wall lists Shopify, Lovable, Canva, Zapier, Apple, Replit, DoorDash, and others.
What you get
Markdown-by-default output is the shape an LLM context window actually consumes. SDKs in Python, Node.js, Go, Rust, Java, and Elixir plus a CLI and REST API. Open source under AGPL-3.0 at github.com/mendableai/firecrawl with a published self-hosting guide at docs.firecrawl.dev/contributing/self-host. P95 latency of 3.4 seconds and a claim of 96% web coverage including JS-heavy pages, both per the homepage (vendor numbers, not independent benchmarks). Official MCP server at github.com/mendableai/firecrawl-mcp-server for Claude Desktop, Cursor, and any other MCP-aware client. Framework integrations published for LangChain, LlamaIndex, Dify, Flowise, CrewAI, Langflow, Camel AI, and SourceSync.ai.
Who should use it
AI builders shipping RAG, agent tools, or research assistants who need reliable URL-to-Markdown ingestion without standing up a Playwright fleet. Solo indie hackers building niche search engines or content aggregators where the Hobby tier ($16/month yearly, 5,000 credits, 5 concurrent per firecrawl.dev/pricing) carries the load. Data engineers at mid-market companies evaluating Firecrawl against an in-house Playwright cluster who want the AGPL-3.0 escape hatch. Skip it if you need raw HTML/DOM access, managed browser automation as a primary use case, or you are already running a Playwright cluster at the volume where you have absorbed the DevOps cost.
Pricing
From firecrawl.dev/pricing, fetched 2026-05-19. Free at $0/month carries 1,000 credits and 2 concurrent requests. Hobby at $16/month (billed yearly) covers 5,000 credits and 5 concurrent. Standard at $83/month yearly is "Most popular" with 100,000 credits and 50 concurrent. Growth at $333/month yearly carries 500,000 credits and 100 concurrent. Scale at $599/month yearly is 1,000,000 credits and 150 concurrent. Enterprise is custom and includes SSO, SLA, zero-data retention. A crawl call consumes one credit per page, so an unbounded firecrawl.crawl() on a doc-heavy site can drain a Hobby month in a single run; teams should pin the limit argument.