Scrape + AI-extract 10,000 pages for $39.
No AI credits. No metered extras.
Firecrawl and Tavily charge per AI extraction. We run inference on our own GPU — so structured JSON costs the same as raw HTML: zero extra.
500 free pages — no card required.
Built for RAG & AI agents · LlamaIndex · LangChain · claude mcp add
Them — metered AI
10,000 pages, AI extraction on each
- 10k page fetches~$30
- + AI extraction credits+ $$$ per call
- Totalsurprise bill
TheCrawler — AI included
10,000 pages, AI extraction on each
- 10k pages (Starter)$39
- + AI extraction$0 — included
- Total$39 flat
One call. URL in, validated JSON out.
No LLM endpoint to wire up — extraction runs on our GPU. Page-grounded, so it copies what the page says and returns null for anything it can't verify. You only pay for pages that succeed.
- 1. POST a URL + a schema or prompt.
- 2. We crawl, render JS, extract on our Qwen GPU, ground it to the page.
- 3. Poll for clean JSON. Failures aren't charged.
POST /v1/extract
{
"urls": ["https://books.toscrape.com/…"],
"extractJsonSchema": {
"title": "string",
"price": "string"
}
}
→ 200 (AI included, $0 surcharge)
{
"title": "A Light in the Attic",
"price": "£51.77",
"nulledFields": [] // grounded
}Every feature kills a metering pain
Unmetered AI extraction
No per-call surcharge — we own the GPU. The +credit AI tax is gone.
Page-grounded
Every value must trace to the page text or it returns null. 0 invented fields on our 29-page trap eval; 1 in 34 on the harder extended run — we report both numbers.
/diagnose
A no-LLM readiness check — we don't charge for extractions that would've failed.
Validated contracts
Built-in real-estate / product / docs schemas with field validation.
Extract-once, emit-all
Markdown + JSON + brand + contacts + HTML in one call. No per-format billing.
Credits never expire
Pay only for successful pages. No expiry tax, no surprise bills.
Flat-rate monthly, or pay per page. AI is free on every page.
A monthly throughput plan for steady production, or a prepaid page pack for bursty use · AI extraction included free on both
Or pay as you go — prepaid pages, never expire
Prices exclude taxes where applicable. TheCrawler is a tool for businesses & developers.
Questions devs actually ask
Will you add AI extraction fees later?
No. The whole model is to NOT meter the AI — you pay for pages, extraction is included. We own the GPU, so there's no per-call cost to pass on.
What counts as a successful page?
A page we fetched and (for extraction) returned a non-empty result for. Fetch failures, empty pages, timeouts, and blocked URLs are not charged — they're auto-refunded.
How is this different from Firecrawl?
Firecrawl charges extra credits for AI extraction on top of page costs. We don't — extraction is free on every page. For extraction-in-a-loop workloads, that's a much lower total bill.
Does it invent data?
It's built not to: extraction is page-grounded — values are verified against the page text the model saw; anything unverifiable is nulled and listed in nulledFields. On our trap evals: 0 invented fields in the original 29-page run, 1 in 34 on the harder extended run (an eval, not a proof) — and failed or empty extractions are auto-refunded.
Do you store my scraped data?
No long-term store. Results live ~24h so you can fetch them, then expire. We don't sell data or train on it.
Drop the meter. Own the crawl.
500 free pages, AI extraction included. No card to start.