What it is
Firecrawl turns web pages into LLM-ready Markdown or JSON through an API that can scrape a page, crawl a site, map its URLs, search the web, parse PDFs and DOCX files, and drive a browser for clicks and form input. The repository moved from mendableai to the firecrawl organisation.
The open-source edition is AGPL-3.0 and covers scrape, crawl, map and search with plain fetch and Playwright. Fire-engine, the anti-bot and browser layer behind screenshots and actions, is not included, and agent and interact are cloud-only.
Firecrawl Cloud is billed in credits: Free gives 1,000 a month, Hobby is $19 for 5,000, Standard $99 for 100,000, Growth $399 for 500,000 and Scale $749 for 1,000,000. A scrape costs one credit; JSON extraction adds four.
Our review
Firecrawl is a commercial API with an open-source core, and it should be judged that way. The cloud is what most of its 194,500 GitHub stars are buying into: clean Markdown, a small set of endpoints and an official MCP server, with JavaScript rendering, proxies and bot walls handled for you. The last tagged GitHub release is v2.11.0 from 19 June 2026, but the main branch moves daily and the Python SDK reached 4.54.0 on 22 October 2026.
Credits are cheap at the bottom and add up in the middle. A plain scrape is one credit, but JSON-format extraction adds four, so structured data from 100,000 pages uses 500,000 credits, the whole Growth plan at $399 a month. Search costs two credits per ten results and browser interaction two credits a minute. Extra credits come in $5 packs whose size depends on the plan, from 1,000 on Hobby to 5,000 on Scale, and only Scale rolls unused credits over, for one month.
Self-hosting is harder than the AGPL label suggests. Firecrawl's own guide says the default Docker stack disables API authentication, is 'not a production architecture' and has no verified minimum host size; it runs Redis, PostgreSQL, RabbitMQ and Playwright workers. Screenshots and actions need Fire-engine, which is not included, and extraction needs your own OpenAI-compatible provider or Ollama. The AGPL also obliges you to publish changes if you offer a modified version over a network.
Security is worth tracking on both sides. In July 2026 the project disclosed a critical remote code execution flaw reachable through /scrape's change-tracking git-diff mode and an attacker-controlled website, alongside a high-severity arbitrary file read via JSON Schema references. For the cloud, Firecrawl says it holds SOC 2 Type II, and offers zero data retention and SAML or OIDC single sign-on on Enterprise.
Crawl4AI is the obvious comparison and now sells a hosted API too, at $10 a month for 10,000 credits against Firecrawl's $19 for 5,000. Firecrawl's edge is maturity, per-plan rate limits in writing and compliance paperwork; Crawl4AI's is a permissive licence and a library that fits inside one Python process.
Where it earns its place
An agent developer adding web access to Claude Code or Cursor
Let the agent search and read live pages
Install the official MCP server with an API key; a scrape is one credit and a search two per ten results. Free's 1,000 credits and 10 scrapes a minute suit a prototype, and Hobby at $19 lifts the limit to 100 a minute.
A growth team at a 50-person e-commerce firm
Track competitor prices across 20,000 product pages every week
Crawl with JSON extraction at five credits a page, roughly 430,000 credits a month, which fits Growth at $399 with 500,000 credits and 50 concurrent browsers. Failed or re-run pages eat into the margin quickly.
A platform engineer at a bank that cannot send pages to a third party
Run Firecrawl inside the network
Deploy the AGPL Docker stack with Redis, PostgreSQL and RabbitMQ, switch on database authentication and add TLS. Expect no anti-bot layer, no screenshots and no official sizing; Enterprise cloud with zero data retention is the managed alternative.
Before you commit
- Setup
- Cloud: sign up, take an fc- API key and call /v2/scrape or install the Python or Node SDK; a first page returns in seconds. Self-host: Docker Compose with Redis, PostgreSQL, RabbitMQ and Playwright, plus your own LLM provider for extraction.
- Integrations
- REST v2 API, Python and Node SDKs, a CLI and an official MCP server; endpoints for scrape, crawl, map, search, batch scrape, parse, monitor, interact and agent.
- Data and privacy
- Cloud: Firecrawl states SOC 2 Type II, GDPR with a DPA, and zero data retention plus SSO on Enterprise. Self-hosted: the default stack has no API authentication and must not be exposed publicly; the self-host guide does not mention telemetry.
- Cost at scale
- Hobby $19 (5,000 credits), Standard $99 (100,000), Growth $399 (500,000), Scale $749 (1,000,000) a month, about 17 to 20% less billed annually. JSON extraction multiplies page cost by five; extra credits cost $5 per 1,000 to 5,000 depending on plan.
- Limits
- Rate limits run from 10 scrapes a minute on Free to 10,000 on Scale, and crawl or agent calls from 2 to 2,000; credits roll over only on Scale; 657 open issues and pull requests; site owners have asked for an identifiable user agent so they can block it.
What's good
- Clean Markdown with scrape, crawl, map, search and parse
- Official MCP server plus Python and Node SDKs
- Published rate limits per plan and a spending cap
- SOC 2 Type II and zero data retention on Enterprise (vendor-stated)
What's not
- Fire-engine, screenshots, actions and agent are cloud-only
- Self-host stack ships without auth or sizing guidance
- JSON extraction costs five credits a page
- Critical RCE advisory in July 2026
Best for
Developers who want a managed API to turn websites into LLM-ready data.
Alternatives
Firecrawl: common questions
- Is Firecrawl worth it?
- A hosted web layer for agents that earns its fee; the open-source edition is a starting point, not a substitute. We rate it 4.2 out of 5.
- How much does Firecrawl cost?
- AGPL-3.0 open source. Cloud: Free 1,000 credits/month; Hobby $19, Standard $99, Growth $399, Scale $749 a month; Enterprise custom.
- What are the downsides of Firecrawl?
- Fire-engine, screenshots, actions and agent are cloud-only; self-host stack ships without auth or sizing guidance; jSON extraction costs five credits a page; critical RCE advisory in July 2026.
- Who is Firecrawl best for?
- Developers who want a managed API to turn websites into LLM-ready data.
- How much does Firecrawl cost for a team or in production?
- Hobby $19 (5,000 credits), Standard $99 (100,000), Growth $399 (500,000), Scale $749 (1,000,000) a month, about 17 to 20% less billed annually. JSON extraction multiplies page cost by five; extra credits cost $5 per 1,000 to 5,000 depending on plan.
- Is Firecrawl safe for company data?
- Cloud: Firecrawl states SOC 2 Type II, GDPR with a DPA, and zero data retention plus SSO on Enterprise. Self-hosted: the default stack has no API authentication and must not be exposed publicly; the self-host guide does not mention telemetry.
- What are the best Firecrawl alternatives?
- The closest alternatives we have reviewed are Crawl4AI, Browser Use and Docling.
Answers checked Oct 23, 2026.
Make Firecrawl? Show this rating on your site
