What it is
Primetra checks how AI assistants and agents see a product's website. PrimetraBot is how it reads the public pages it needs for that. Every request says so in its user agent:
PrimetraBot/1.0 (+https://primetra.ai/bot)What it fetches
- A small number of public pages on a site a customer, or someone running our free check, asked us to look at: the homepage, pricing and docs pages, and the sitemap.
- The files agents read on that site: robots.txt, llms.txt, and API and MCP description files.
- Lists that AI companies publish of their crawlers' addresses and signing keys, so we can tell their real agents from impostors.
- An assistant's own redirect links in its answers, to learn which page it cited. We read where the link points, not the page.
It only reads. It sends GET and HEAD requests to public addresses, never fills in a form, never signs in, and stops after a few seconds or a few megabytes.
robots.txt
PrimetraBot reads a site's robots.txt before it fetches a page there, and follows the rules for PrimetraBot, or the rules for * when there are none for it. No robots.txt (a 404) means no rules. If robots.txt can't be reached (a server error or a timeout), it fetches nothing from that site. The copy it works from is at most an hour old, so a change reaches it within the hour.
Some fetches don't consult robots.txt: robots.txt itself, the address and key lists and redirect links above (their owners publish them for exactly this use), and, for a paying customer, a few pages of their own site that we read to draft the fixes they asked for.
How to block it
Add these two lines to your robots.txt. PrimetraBot then fetches nothing more from your site, apart from robots.txt itself (and, if you're a customer, the pages we read to draft your fixes).
User-agent: PrimetraBotDisallow: /Contact
Questions, or a problem with how PrimetraBot behaves on your site: email hello@primetra.ai. For a security issue, email security@primetra.ai.