Kupeta's agent
If you found Kupeta in your logs, this page is for you: who visits, what it does on your website, what it never does, and how to keep it out.
Last updated 19 November 2026
1Who visits
Kupeta tests whether AI shopping assistants can buy from online stores. To do that, it opens websites in an ordinary browser, the way an AI assistant would. A visit comes either from a free test, which anyone can start for any store, or from a full test, which only the verified owner of a website can start.
Kupeta always says who it is. It never disguises itself as a person or hides that the browser is automated.
2How to recognize it
Every request carries Kupeta's name at the end of the browser's own User-Agent:
Kupeta/1.0 (+https://kupeta.example/bot)In robots.txt, Kupeta reads the rules for the user agent Kupeta, and if there are none, those for *.
Signed requests by the Web Bot Auth standard (HTTP Message Signatures) are prepared and will be switched on with our own domain. This page will then name the address of our key directory.
3What a free test does
A free test only looks around. It reads:
- robots.txt, first, and nothing else if it asks Kupeta to stay away
- the sitemap, if there is one
- the homepage and one product page
It doesn't add anything to a cart, fill in forms, log in, create accounts or send anything to your website. Screenshots and page text are saved only after personal data in them is covered.
4What a full test does
A full test runs only on a website whose owner verified it in Kupeta (with a meta tag, a file or a DNS record) and agreed to it. Test shoppers then go through real purchases: they choose a product, put it in the cart and fill in the checkout with made-up details.
They stop before the last confirmation, whatever the button says. Nothing is ever ordered or paid, and card details are never entered. The store may see abandoned carts from these tests.
5How gently it goes
- At least one second between pages, longer if your robots.txt asks for it with Crawl-delay.
- At most 300 requests in one look around your website, images and scripts included.
- A captcha or a bot wall ends the attempt. Kupeta never tries to get around it: it's reported to the store as a finding.
6How to keep Kupeta out
Add this to your robots.txt. From the next test on, Kupeta reads robots.txt and opens no other page of your website:
User-agent: Kupeta
Disallow: /You can also close only some parts of the website with Disallow rules for those paths. Kupeta follows the most specific rule, as robots.txt is meant to be read (RFC 9309).
A report of a website that asks Kupeta to stay away says so, and no shopper is sent there.
7Questions
To be added: The contact address for website owners will be added here.