Describe what you want. Get it back.
Plain-English prompts → structured data from any page. No schema required, no DOM walking, no fragile parsing. The slowest, most flexible tool in the kit, for the long tail of weird pages.
POST /ai-scraper
{
"url": "https://news.example.com/2026/may/01/agents-eat-saas",
"prompt": "extract title, author, publish date, and a 2-sentence summary"
}
→ 200 OK · 25 credits · 2.4s
{
"title": "Agents eat SaaS",
"author": "L. Reyes",
"publish_date": "2026-05-01",
"summary": "A look at how autonomous agents
are dissolving the line between buyer and
user. Implications for product pricing."
}See it work, before you sign up.
Drop in a URL, run a real call against the live API, and watch the JSON come back in about a second. No API key required.
Skip the JSON schema dance. Describe the fields you want in English; we'll return them named and typed. Add examples for stricter shape.
When AI Scraper's built-in schemas don't fit (odd directories, government portals, niche listings), AI Scraper handles them without per-site engineering.
Pass cite: true and every field comes with the source DOM path it was lifted from. Auditable. Debuggable. No "trust the model".
For long pages, set stream: true and we emit fields as they're extracted. Show partial results in your UI before the full call returns.
Tell it to follow links, paginate, or scroll-load. AI Scraper plans the trajectory, executes steps, and returns the merged result.
Bring your own OpenAI/Anthropic/Mistral key with byo_llm: true and we just handle fetching + parsing. We never train on your data.
You don't set a proxy_type. AI Scraper escalates through the same auto_proxy ladder as the Scrape API (datacenter → residential → premium → managed-Selenium) and we only bill the tier that returns 200. CAPTCHA solving is on by default too. The flag exists if you want to pin it; you almost never need to.
Copy. Paste. Ship.
import { Ujeebu } from "ujeebu";
const uj = new Ujeebu(process.env.UJEEBU_KEY);
const { data } = await uj.ai({
url: "https://docs.weird-niche-portal.gov/proc/3801",
prompt: "extract case number, filing date, parties, and outcome",
cite: true
});from ujeebu import Ujeebu
uj = Ujeebu(api_key=os.environ["UJEEBU_KEY"])
result = uj.ai(
url="https://docs.weird-niche-portal.gov/proc/3801",
prompt="extract case number, filing date, parties, outcome",
cite=True,
)curl -X POST https://api.ujeebu.com/ai-scraper \
-H "Authorization: Bearer $UJEEBU_KEY" \
-d '{
"url": "https://docs.weird-niche-portal.gov/proc/3801",
"prompt": "extract case number, filing date, parties, outcome",
"cite": true
}'const stream = await uj.ai({
url: "https://...",
prompt: "extract all 50 line items as { name, qty, price }",
stream: true,
});
for await (const field of stream) {
console.log(field); // { name: "Linen Arc Lamp", qty: 2, price: 189 }
}Real things real teams shipped this quarter.
Government procurement portals, university course catalogues, niche industry directories: sites no schema covers. AI Scraper gets you running in 5 minutes.
Pass a paper PDF or news article and a domain-specific prompt ("extract every named clinical trial and its NCT ID"). Returns structured rows.
Pages where you have to fill a search form to see anything useful. AI Scraper plans the interaction and extracts the post-submit data.
Heterogeneous result pages (real estate + auctions + classifieds in one feed). The model sorts by item type and emits per-type structured rows.
25 credits per AI Scraper call. Here's what that buys.
One credit pool covers every endpoint. Failed calls cost 0. No per-feature upcharges, no premium-proxy tax. See full pricing →
Ship AI Scraper tonight.
5,000 credits free. No card. Real residential proxies on the free tier.