Web Data and Document Extraction
Firecrawl
A web data API that searches, scrapes, and crawls sites into clean markdown and structured output for AI agents.
Quick decision
- Best for
- Agents and pipelines that need clean markdown or structured JSON from many pages without running browsers themselves.
- Not ideal for
- Fully local, no-account workflows. The practical path is the hosted API, and the open-source stack is a larger self-host than a single library install.
- Pricing model
- Open-source code is available under AGPL-3.0, and a hosted service is sold at firecrawl.dev. AgentsUse does not list hosted prices here because they change. Check the official pricing page for current plans.
- Deployment
- Hosted API. Self-hosted (open source).
- Authentication
- Hosted API uses a Firecrawl API key (FIRECRAWL_API_KEY). Keep the key in your environment and use a placeholder in examples.
- Review state
- Source verified. Facts checked 2026-10-03. Not locally tested by AgentsUse.
What it does
Firecrawl turns web pages into LLM-ready data. Its core endpoints search the web, scrape a URL to markdown or structured JSON, crawl a whole site, map a site URL list, and batch-scrape many URLs.
The project is open source under AGPL-3.0 and is also offered as a hosted service at firecrawl.dev. SDKs cover Python, Node.js, Go, Java, Rust, and more, and an MCP server (firecrawl-mcp) connects MCP clients to the API with a Firecrawl API key.
An Agent endpoint gathers data from a prompt without requiring URLs up front, with optional structured output through a schema. Hosted use needs an API key from firecrawl.dev.
Verified capabilities
Search and retrieval
Web search with content Source verified
Search the web and get full page content from the results.
Data access
URL to markdown or JSON Source verified
Scrape a URL to clean markdown, HTML, screenshots, or structured JSON.
Site crawl and map Source verified
Crawl all URLs of a website or discover its URL list with a map request.
Batch scrape Source verified
Scrape thousands of URLs asynchronously.
Agent integration
Prompt-driven agent endpoint Source verified
Describe the data you need and let the hosted agent search, navigate, and retrieve it, with optional structured output.
MCP server Source verified
firecrawl-mcp connects MCP clients to the API using your Firecrawl key.
Quick start
Call the hosted API
Shape from the official README quick start. Sign up at firecrawl.dev for a key, and replace the placeholder with your own key in your environment.
curl -X POST https://api.firecrawl.dev/v2/scrape \
-H "Authorization: Bearer YOUR_FIRECRAWL_API_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://example.com"}'Environment variables (names only, use your own values)
FIRECRAWL_API_KEY
Source: https://github.com/firecrawl/firecrawl. Examples use placeholders only. Never paste a real key into a profile, config file you share, or a ticket.
MCP support: Yes. The firecrawl-mcp package connects MCP clients to Firecrawl with a FIRECRAWL_API_KEY, per the project README.
Works with
MCP
firecrawl-mcp package (npx), API key required.
Packages
Python and Node.js SDKs, plus Go, Java, Rust, Ruby, .NET, and PHP per the README.
API
Hosted REST API at api.firecrawl.dev.
Deployment
Hosted API, or self-host the AGPL-3.0 code.
Only sourced support is listed. A missing framework means AgentsUse has not verified it yet, not that it cannot work.
Health and maintenance
- GitHub stars
- 188,305 (checked 2026-10-03)
- GitHub forks
- 10,018 (checked 2026-10-03)
- License
- Open source, AGPL-3.0
- Maintainer
- Firecrawl
Stars and forks from the GitHub repository page. AgentsUse does not show a package version here because the API and SDK versions move separately. Maintenance signals only, not a quality rating.
Pricing and license
Open-source code is available under AGPL-3.0, and a hosted service is sold at firecrawl.dev. AgentsUse does not list hosted prices here because they change. Check the official pricing page for current plans.
Limitations and safety
- The hosted API needs an account and a key, and usage is metered by the vendor. Check current pricing before building a high-volume job on it.
- The open-source license is AGPL-3.0. Review what that means for your deployment before self-hosting or embedding it.
- No independently checked benchmark is published by AgentsUse for this tool yet. Coverage and latency figures on the vendor site are vendor claims.
Browser and data tools can read pages, fill forms, and download files. Start with a test account or read-only access, keep credentials in environment variables, and review agent actions before connecting anything that can spend money, send messages, or delete data.
Alternatives to Firecrawl
Choose Crawl4AI when you want a Python library you run yourself, free, with deep crawl and extraction strategies.
Tradeoff: You run the browsers and infrastructure. There is no hosted search API in the library path.
Choose Jina Reader when you want the simplest URL-to-markdown call, with a prefix and no SDK.
Tradeoff: Less crawl and batch machinery. Heavy jobs need the hosted API limits checked first.
Choose MarkItDown when your input is local files and Office documents rather than live web pages.
Tradeoff: It converts files you already have. It does not crawl or search the web.
Related tools
Common questions
What does Firecrawl do for an AI agent?
A web data API that searches, scrapes, and crawls sites into clean markdown and structured output for AI agents.
Is Firecrawl open source?
Yes. This profile records the license as AGPL-3.0 from the official repository.
Does Firecrawl support MCP?
Yes. The firecrawl-mcp package connects MCP clients to Firecrawl with a FIRECRAWL_API_KEY, per the project README.
Sources and freshness
- GitHub repository: https://github.com/firecrawl/firecrawl
- Documentation: https://docs.firecrawl.dev
- Product site: https://www.firecrawl.dev
Last checked 2026-10-03. Verification label: source verified, which means public claims trace to the sources above. It does not mean AgentsUse ran the tool. Spotted an error? Send a correction. Back to the tools directory.