Our bots
Last updated 3 November 2026
enivei gives websites an AI assistant, and a connection that lets ChatGPT, Claude and other assistants use them. To answer people's questions, two bots of ours read websites, and a third while we build a website's assistant. This page says what each one does, how to recognise it and how to let it in.
What each bot does
EniveiBot
Reads a website's public pages ahead of time, so that its assistant can search them: help, delivery and contact pages, for example. For a shop whose price history we keep, it reads the product listings. It also asks a few questions the way a person's assistant would, to check that the connection still works.
Follows robots.txt: the rules for EniveiBot, or else those for all bots, and Crawl-delay.
Mozilla/5.0 (compatible; EniveiBot/1.0; +https://enivei.com/bots)
Enivei-User
Looks something up when a person asks an AI assistant a question that needs the website, such as a product's price or whether it is in stock. It fetches what that question needs and does not crawl. For a website set up on enivei, it also reads that website's home page, for its own assistant.
Follows the robots.txt rules written for Enivei-User by name. Rules for all bots (User-agent: *) don't apply to it, because a person asked. Reading the home page for a website's own assistant doesn't check robots.txt: it's done for that website.
Mozilla/5.0 (compatible; Enivei-User/1.0; +https://enivei.com/bots)
Enivei-Build
Reads a website while we build its assistant, when the website's owner asks for one: its pages, in a browser and directly, for as long as the build takes. We also use it when we test an assistant.
Doesn't ask robots.txt for permission: the website's owner asked for the build.
Mozilla/5.0 (compatible; Enivei-Build/1.0; +https://enivei.com/bots)
None is used to train AI models.
When a website says no
When a website refuses a request, limits how often we may ask or shows a challenge, we take that as its answer. We don't retry from other addresses, pretend to be a browser or solve CAPTCHAs.
How to recognise them
- The User-Agent names the bot, as above.
- The address of EniveiBot and Enivei-User is one of these: 49.12.235.39. The same list for your firewall to read: enivei.com/bots.json.
- Enivei-Build comes from our build provider's cloud, whose addresses others use too, so only its User-Agent tells it apart.
How to let them in
Bot protection may stop our bots before they reach your website, and its assistant then cannot answer. To let them through, allow our addresses, not the User-Agent alone, since anyone can copy a User-Agent.
- Cloudflare with Bot Fight Mode (the free plan): add an IP Access rule that allows each address above. Bot Fight Mode takes no other exception.
- Cloudflare with Super Bot Fight Mode: add a custom rule with the action Skip, skipping Super Bot Fight Mode, with this expression:
ip.src in {49.12.235.39} - Any other firewall: allow the addresses above, or have it read enivei.com/bots.json.
To let a build of your assistant through, allow the User-Agent Enivei-Build. Anyone can copy a User-Agent, so take the rule away once your assistant is ready.
If your robots.txt keeps all bots out, EniveiBot obeys it. Let it in by name, for your whole website or a part of it:
User-agent: EniveiBot Allow: /
Enivei-User needs no robots.txt rule, as it follows only rules that name it. We read robots.txt again at least once a day, so a change reaches EniveiBot and Enivei-User within a day. To keep one of them out instead, write Disallow: / under its name.
Contact
If our bots cause trouble on your website, or you have a question, write to [email protected] from an address at your website's domain, so that we can tell it is yours. enivei is Enivei Technology OÜ, Estonia.