Firecrawl turns websites into clean data that an AI can use. Give it a URL and it returns the page as Markdown, JSON or a screenshot, after dealing with JavaScript, pop-ups, anti-bot checks and messy HTML. Give it a query and it searches the web and scrapes the results in the same call.
Since September 2026 it also sells data it doesn't scrape: Alexandria, a catalogue of 100+ data providers behind the same API key. That's how Firecrawl now returns people and company data for a LinkedIn URL.
| Company | Firecrawl, San Francisco, from Y Combinator |
| Founders | Caleb Peffer, Eric Ciarla and Nicolas Silberstein Camara, who built Mendable before |
| Funding | $14.5M Series A in August 2025, $75M Series B in September 2026 led by Smash Capital |
| Open source | firecrawl/firecrawl, AGPL-3.0, about 187,000 GitHub stars |
| Access | REST API, Python and Node SDKs, CLI, MCP server for Claude and other agents |
| Security | SOC 2 Type II, PII redaction, zero data retention on Enterprise |
What it does
| Endpoint | What it does | Credits |
|---|---|---|
| Scrape | 1 page to Markdown, HTML, links, screenshot or JSON | 1 per page |
| Crawl | Follows a site's links and scrapes every page | 1 per page |
| Map | Lists every URL on a site, very fast | 1 per page |
| Search | Searches the web, optionally scrapes the results | 2 per 10 results |
| Interact | Drives a browser: clicks, forms, logins | 2 per browser minute |
| Monitor | Checks pages for changes on a schedule | 1 per page per check |
| Agent | Researches a question across sources on its own | Preview, 5 free runs a day |
JSON extraction and question formats add 4 credits per page. A scrape that returns nothing isn't charged, but a 403 or 404 page is.
Alexandria
Alexandria launched with the Series B on 23rd September 2026. It puts 100+ data providers and Firecrawl's own indexes behind 1 API: Apollo, FullEnrich, DataLegion, financial and public-record sources, research papers and more. An agent asks for a capability, such as "company enrich" or "work email", and Firecrawl calls the provider and bills it on the same credits. Firecrawl says agents using Alexandria score 21% higher on answer quality.
LinkedIn enrichment
30 Sep 2026 Firecrawl now returns people and company data for a LinkedIn URL, from the providers in Alexandria. The data comes from the providers' own databases, not from scraping LinkedIn.
To turn it on:
- Open the Alexandria enrichment settings.
- Pick the providers to use, and the order to try them in.
- Accept the providers' terms.
From then on, every endpoint enriches the LinkedIn URLs it meets. The providers I'd look at:
- FullEnrich: waterfall enrichment across many data vendors. Through Alexandria it finds a verified work email or a mobile number from a LinkedIn URL, and bills only when it finds a result.
- DataLegion: people and company matching, with and without contact details.
- Apollo: company data from a domain or a LinkedIn company URL.
For business people
The problem: websites are built for people, not for software. Every team that feeds web content to an AI rebuilds the same scraper, and it breaks every time a site changes.
What Firecrawl does: 1 API that returns clean content from any page, and now business data from providers, on 1 bill.
Where it fits: account research before a call, competitor and pricing monitoring, lead lists from public directories, and giving an AI agent live web access.
Pricing
| Plan | Price (billed yearly) | Credits a month |
|---|---|---|
| Free | $0 | 1,000 |
| Hobby | $16 a month | 5,000 |
| Standard | $83 a month | 100,000 |
| Growth | $333 a month | 500,000 |
| Scale | $599 a month | 1,000,000, with 1 month of rollover |
Paid plans top up in $5 steps when credits run out, up to a monthly limit you set. Alexandria providers bill on their own terms, shown before you accept them.
The limits:
- No site is scraped perfectly. Anti-bot systems and heavy apps still fail sometimes.
- Credits add up. JSON extraction costs 5 credits a page, not 1, and crawls grow fast.
- Enrichment is personal data. In Europe, GDPR applies to what you store and how you use it, whoever provided it.
For technical people
A scrape with the v2 API:
curl -X POST https://api.firecrawl.dev/v2/scrape \
-H "Authorization: Bearer $FIRECRAWL_API_KEY" \
-H "Content-Type: application/json" \
-d '{"url": "https://www.firecrawl.dev/pricing", "formats": ["markdown"]}'
- Formats:
markdown,html,rawHtml,links,screenshot,summary,jsonwith a schema,queryfor a direct question, andbranding. - Freshness: Firecrawl can serve a recently cached copy. Set
maxAge: 0for a live fetch. - Alexandria calls go through the same
/v2/scrapeendpoint with analexandriaarray of{provider, capability, options}, up to 10 per request. Each call can fail on its own, so check each result. - MCP: the hosted MCP server gives Claude Code search and scrape without a key, with limits. An API key unlocks map, crawl, agent and the Alexandria catalogue.
- Self-hosting: the core is open source under AGPL-3.0, but the hosted API has the anti-bot and browser infrastructure.
My usage
14 Feb 2026 started using Firecrawl.
It's the scraping layer of my whole setup. My shared script firecrawl.py turns any URL into Markdown for research, call prep and account briefs, and my image harvester uses it as the fallback when a page hides its images. Claude Code also reaches it through the MCP server.
30 Sep 2026 added contact enrichment. After testing FullEnrich against DataLegion, I now use FullEnrich through Firecrawl for LinkedIn URLs: 1 API key and 1 bill for both scraping and contact data. For my other data vendors, see Data providers for B2B sales lead generation .
Usage dashboard: firecrawl.dev/app/usage.
Further Reading
