Describe the data. Skip the selectors.
Add `ai_query` to a crawl and Irmu answers a question about the page once it has been retrieved. The response is the JSON envelope with the answer in `data` rather than a document you have to parse, and layout changes stop breaking your pipeline overnight.
What AI Extract handles for you
Plain-language questions
Describe what you want in a sentence — no selectors, no schema to maintain.
Runs on the fetched page
One request fetches and answers, so you are never holding HTML you do not want.
Boilerplate removed first
Ads and trackers are blocked before the page loads, so the model reads the content.
Bring your own model
Configure a model in organization settings and AI questions cost no additional credits.
Priced per question
5 credits on top of the fetch. An empty ai_query costs nothing extra.
Same contract as a crawl
Auth, quota, 202 polling and billing behave exactly as they do for any other crawl.
Request
curl -G https://app.irmu.com/api/crawl \
-H "Authorization: Bearer $IRMU_API_KEY" \
--data-urlencode "url=https://www.amazon.com/dp/B0CHX1W1XY" \
--data-urlencode "ai_query=Return the product title, price and currency as JSON"Response
{
"url": "https://www.amazon.com/dp/B0CHX1W1XY",
"domain": "amazon.com",
"premium": false,
"js": true,
"ai_query": "Return the product title, price and currency as JSON",
"credits_charged": 10,
"credits_remaining": 990,
"status": "completed",
"data": {
"title": "Kindle Paperwhite (16 GB)",
"price": 149.99,
"currency": "USD"
},
"error": null
}Measure it on your own targets
We don't publish benchmark tables, because latency and success rates depend entirely on the sites you crawl. The free plan is there so you can measure AI Extract against your real target list before you pay for anything.
Run your first request →Credits per operation
- 1 credit
Standard fetch
Datacenter IP, no JavaScript execution
- 5 credits
Standard fetch + JavaScript
Page rendered in a real browser
- 10 credits
Premium fetch
Residential IP, rarely blocked
- 25 credits
Premium fetch + JavaScript
Residential IP and full rendering
- 5 credits
AI question about a page
Free with your own language model
- Free
Country targeting
Included on every request
AI Extract questions
Other products
Web Scraping API
Fetch any page over GET /crawl, with real browser rendering and residential IPs.
Browser API
Scripted browser sessions — click, type, wait. Not available yet.
Screenshot API
Full-page captures over GET /screenshot, returned as base64 JPEG.
Proxy API
Point your own crawler at Irmu's IP pool. Not available yet.
Start building with Irmu today
200 free credits every month, no card required. Every API, every integration, one key.