Paste any page URL, name the attributes you want - price, description, rating, whatever the page carries - and get them back as columns. There is no per-site parser to wait for: the page is read and the fields you asked for come out as a row.
Two inputs and nothing else: where to look, and what to bring back. The second one is the part that makes this work on a site nobody has written a parser for.
Each page is fetched through our proxy pool, never your real IP, and read by Claude, which pulls out the attributes you named.
The attributes you type are the schema. Ask for a price and a rating and you get two columns; add a description and you get three. Nothing here has a fixed header row, which is exactly why it fits pages that differ from each other.
Most sites will never be popular enough to get a purpose-built scraper. This is the answer for those: a supplier catalogue, a trade directory, a regional marketplace - anything with the data on the page but nobody maintaining an extractor for it.
However you define the columns, the output is the ordinary CSV, XLSX or JSON the rest of the platform produces. It drops into the same pipeline as a run from any of our fixed-schema scrapers.
Every other page on this site names a fixed set of columns. This one cannot, because you write it.
The attributes you add are the output. Type price and rating and the file has two columns with those names. The picker suggests a few common ones to get you started - name, price, discount price, description, image, rating, reviews - but the field takes free text, so the list is a convenience rather than a limit. Ask for whatever the page actually carries.
That flexibility is the whole product, and it comes with an obvious consequence: we cannot show you a sample export or promise what any given cell will hold, because both depend on the page you point at and the words you choose. Name attributes the way the page itself labels them, expect an empty cell when a page simply does not carry something you asked for, and prove the wording on a couple of pages before you turn it loose on a list.
One practical difference from the rest of the platform: the pages limit starts at 1, not 100. That is a sensible default for a tool you are meant to try on one page first, but it does mean a run will stop after a single page until you raise it.
There is no supported-site list to check against, because there is no per-site parser. The two required inputs are the URLs and the attributes you want out of them.
A few examples of what a scraper with no fixed schema is actually for.
A regional marketplace, a trade association directory, a manufacturer's catalogue. The data is on the page and there is no dedicated tool for it - name the fields you need and take them.
Point it at a handful of pages with a rough attribute list and see whether the data you need is even there. That is a cheaper first step than specifying a bespoke extractor for a source you have not validated.
Sometimes a dedicated scraper returns almost everything you want. Point this at the same pages, ask only for the missing attribute, and join the result back on the URL.
No subscription, no minimum, no recurring bill. Your first 500 rows are on us - after that, pay-as-you-go at the same flat rate as every other scraper here.
Every new account, one-time. No credit card required. Custom attribute lists, file upload and every export format included.
The same flat rate as the rest of the platform. The pre-flight estimator shows the row count and credit cost before a run starts - no surprise bills, no compute units to translate.
Volume pricing, dedicated workers and an SLA for running large page lists on a schedule. Tell us your numbers and we will quote.
LIVESCRAPER10 at checkout.This tool exists for sites that have no purpose-built extractor. Where one does exist, it is the better choice - it knows the site's quirks and ships a documented column set.
For local businesses, use this rather than describing the fields yourself. It returns a documented column set and handles the pagination and place-id details a general tool would not know about.
The same idea with more control over the output. Instead of picking attributes you write a prompt, or build a schema field by field with types and required flags, and its limit counts rows per URL rather than pages per query. Reach for it when the shape of the result has to stay the same between runs.
If contact details are all you are after from an arbitrary website, this is the narrower and more direct tool. Reach for the general one when you need fields it does not cover.
Short answer: that depends entirely on the page - which is the honest answer for a tool with no fixed target, and it puts more of the judgement on you than our other scrapers do.
Every other scraper here points at one site, so we can tell you what it collects and what that site's terms say. This one points wherever you send it, so we cannot make that assessment for you. The run fetches the page as an ordinary visitor would, through our proxy pool and never your real IP, and reads what is on it - it does not log in, defeat a paywall, or reach anything a visitor could not see. Beyond that, whether a given page may be collected from is a question about that page's terms and your purpose, and it is yours to answer before you point the tool at it.
Be deliberate about the attributes you ask for. A general extractor will return whatever you name, including things that are personal data - a person's name, an email address, a phone number. Asking for them turns an ordinary page-fetch into processing personal data, and in the EU and UK that brings the GDPR with you as the controller: a lawful basis, transparency about where the data came from, and honouring objections and erasure. The tool will not stop you; the obligation is yours.
We run no third-party trackers on the data layer, and your exports auto-delete after 30 days. If you are unsure about a source or a field, take advice before you scale rather than after.
The questions we hear most. Anything else? Talk to us - humans, not bots, write the answers.
Your first 500 rows are free - no card, no subscription. After that it is $0.002 per row, flat.
Most scraping tools are built one site at a time. Somebody studies a page's markup, writes an extractor, and publishes a fixed set of columns - which works beautifully for the handful of sites big enough to justify the effort, and not at all for everything else. The Universal AI-Powered Scraper is for everything else. You give it page URLs and a list of the attributes you want, each page is fetched through our proxy pool and read by Claude, and the attributes you named come back as the columns of an ordinary CSV, XLSX or JSON file.
That is why this page has no data dictionary. Every other product page on this site publishes a fixed header row; this one cannot, because you write the schema at run time. Type price and rating and your file has two columns with those names. The picker suggests a handful of common attributes - name, price, discount price, description, image, rating, reviews - but the field takes free text, so ask for whatever the page actually carries.
The flexibility has a cost worth stating plainly. Because the target and the schema are both yours, we cannot show you a sample export or promise what any given cell will contain, and we publish no accuracy figure. The practical advice is the same as it would be for a colleague: name attributes the way the page labels them, expect an empty cell where a page simply does not carry what you asked for, and validate the wording on two or three representative pages before you point it at a long list. Note too that the pages limit starts at 1 rather than the 100 our other scrapers default to.
Where a dedicated scraper already exists for your source, use it - it knows that site's quirks and ships a documented column set. This one earns its place on the sites nobody has built a parser for, and probably never will. Start free: your first 500 rows cost nothing and need no credit card, and after that it is $0.002 per row, flat.