Firecrawl Pricing Teardown
Continuing a long ago forgotten blog series on pricing of open source products
Hey, it’s Vlad, founder of Beton
I’ve had a habit of writing on pricing of COSS (commercial open source software) companies. Want to bring it back to life.
First one is Firecrawl.
Firecrawl has one of the smartest GTM setups I’ve seen in open source. The trick isn’t in the pricing page — it’s in the licensing.
This post is a part of series on commercial open source software pricing. See full list of articles here.
What is Firecrawl
Firecrawl is a web scraping API that turns websites into LLM-ready markdown or structured data.
Think of it as the plumbing layer between “the internet” and your AI agent — it handles proxies, JavaScript rendering, anti-bot bypasses and gives you clean output.
They’ve raised from Y Combinator and are growing fast in the AI tooling space.
The licensing play
This is where it gets interesting. Firecrawl doesn’t have one product — it has two, with very different licenses.
The core scraping engine is AGPL-3.0
You can self-host it, sure.
But AGPL means that if you wrap it into your product and serve it to users, you have to open source your entire codebase.
So competitors can’t just take Firecrawl and build their own UI on top. And even if you self-host it just for internal use — good luck juggling rotating proxies, headless browsers and anti-bot detection at scale.
That’s the hard part. That’s what you’re actually paying Firecrawl for when you buy their cloud.
Fire Enrich (and many other Firecrawl-based tools) are MIT
This is their data enrichment tool — 650+ stars, fully open, do whatever you want with it. You could fork it and build a competing product tomorrow.
But here’s the thing: Fire Enrich is basically a collection of small system prompts that call GPT and tell it to go scrape the web for company data.
Each “agent” is a thin wrapper.
So when you use it, you end up paying Firecrawl for scraping infrastructure and OpenAI for tokens.
The MIT license is generous because the product is a funnel, not a moat. Smart. Really smart.
Pricing structure
Firecrawl uses credit-based pricing: 1 page scraped = 1 credit.
Scrape & Crawl plans:
Free — 500 credits, no credit card required. Enough to test
Hobby — $16/month for 3,000 credits (~$0.005/page)
Standard — ~$83–99/month for 100,000 credits + 50 concurrent browsers
Growth — $333/month for 500,000 credits + 100 concurrent browsers
Enterprise — custom pricing, sales call
But here’s what catches people off guard: the AI extract feature is a completely separate subscription.
It runs on tokens, not credits. Starts at $89/month.
So if you thought your Standard plan covers everything — it doesn’t. You need two subscriptions to use the full product.
Extract plans (separate):
Starter — $89/month
Explorer — $359/month
Pro — $719/month
Lower-tier plans also cap crawls at 50 pages, which gets painful fast if you’re scraping anything with depth — docs, catalogs, archives.
Does it make sense to pay?
Almost certainly yes, if you’re building AI agents that need web data.
Self-hosting Firecrawl is technically possible but practically painful. Web scraping at scale is an infrastructure nightmare — you need proxy rotation, browser pools, retry logic, anti-detection.
This is the kind of thing where “just run a Docker” doesn’t cut it.
Unlike ClickHouse or Metabase where the self-hosted version is genuinely usable, Firecrawl’s value is the managed infrastructure.
The real question is whether the cost per contact makes sense for enrichment use cases.
When you use Fire Enrich, you’re not buying a database lookup at fractions of a cent. You’re scraping live web pages and running LLM inference on every single query.
That adds up fast — scraping credits + OpenAI tokens per contact.
For high-volume enrichment (think Clay or Apollo scale), this could get expensive compared to traditional data providers that pre-crawl and cache everything.
Firecrawl’s bet is that freshness beats cheapness. Instead of stale database records, you get real-time data every time.
Whether that tradeoff works depends on your volume and how much you care about data being current vs. data being cheap.
This post is a part of series on commercial open source software pricing. See full list of articles here.
Awesome post, thx! I’m using firecrawl almost everyday and didn’t realize that