Mistral AI · User-triggered fetch
MistralAI-User: Mistral's assistant fetcher, and how to control it
MistralAI-User is the agent Mistral AI uses when a person asks its assistant, Le Chat (named Vibe in Mistral's current docs), a question: it may visit a web page to help answer and link it as a source. Mistral says it does not crawl automatically or collect training data. Mistral says the token governs which sites these requests can reach, so a robots.txt rule controls it.
- robots.txt token
MistralAI-User- User agent
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots)- Operator
- Mistral AI
- Type
- User-triggered fetch
- What it feeds
- Live page visits when a user asks Mistral's assistant a question, so it can answer and cite the page; it does not crawl automatically or collect AI training data.
- Follows robots.txt
- Yes. Mistral's docs say MistralAI-User governs which sites these user requests can be made to, which implies robots.txt rules for it are honored. There is no explicit 'we honor robots.txt' sentence and no user-initiated exemption.
- How to verify it
- Mistral publishes the IP ranges MistralAI-User uses (list)
Should you block MistralAI-User?
Blocking MistralAI-User means Mistral's assistant cannot fetch or cite your pages when someone asks about them. It isn't a training crawler, so blocking it gains no training opt-out; for that, Mistral documents a separate MistralAI-Training token.
User-agent: MistralAI-User
Disallow: /
User-agent: MistralAI-User
Allow: /
Add these lines to /robots.txt at your site root. A group naming MistralAI-User replaces your User-agent: * rules for it, so repeat any paths you also want it to skip.
Live check
Is MistralAI-User allowed on your site?
MistralAI-User and Mistral's other crawlers
Mistral's crawler documentation lists three tokens. This page covers MistralAI-User; the other two are mentioned here so you can write a complete rule set.
| Token | What it does | IP list |
|---|---|---|
| MistralAI-User | Visits a page when a user asks the assistant a question, and links it as a source | mistralai-user-ips.json |
| MistralAI-Index | Search indexing | mistralai-index-ips.json |
| MistralAI-Training | Training | None published |
The assistant is Le Chat, named Vibe in Mistral's current docs. If you read an older copy of the crawler page or a search snippet, you will see "Le Chat"; it is the same agent.
What blocking MistralAI-User does, and what it doesn't
When someone asks Mistral's assistant about your product and it decides a live page would help, MistralAI-User fetches that page and links it as a source. Block it and the assistant cannot fetch or cite your pages for those questions, so it answers from what it already has.
Blocking it does not opt you out of training. Mistral says MistralAI-User does not crawl automatically or collect training data; training has its own token, MistralAI-Training. Blocking it also doesn't touch MistralAI-Index.
How robots.txt applies
Mistral's docs say MistralAI-User governs which sites user requests can be made to. Unlike OpenAI's ChatGPT-User and Perplexity's user fetcher (see PerplexityBot), there is no exemption for user-initiated requests. The page doesn't contain an explicit "we honor robots.txt" sentence, so treat compliance as Mistral's stated intent rather than a tested guarantee. That puts it closer to Claude-User.
To block training but keep the assistant able to read and cite you:
User-agent: MistralAI-Training
Disallow: /
User-agent: MistralAI-User
Allow: /
A named group replaces your User-agent: * rules for that bot, so a site-wide wildcard block will catch MistralAI-User unless you give it its own group. Each subdomain needs its own /robots.txt.
Checking that a visit is really MistralAI-User
The documented string is Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots). Anyone can send that, so check the source IP against mistralai-user-ips.json.
The live check above sends a request with the MistralAI-User token and compares it with a browser request, which catches a CDN or firewall block. It doesn't come from Mistral's IPs, so IP-keyed rules may differ. The robots.txt checker shows how your file treats each Mistral token.
Questions
- Is MistralAI-User the bot behind Le Chat?
- Yes. Mistral's crawler page describes it as the agent that visits pages when users ask the assistant a question. The current docs call the product Vibe; older copies of the same page say Le Chat.
- Does blocking MistralAI-User stop Mistral training on my site?
- No. Mistral says MistralAI-User does not collect training data. Mistral documents a separate MistralAI-Training token for training.
- Does MistralAI-User follow robots.txt even though a user triggered it?
- Mistral says the token governs which sites user requests can be made to, and gives no user-initiated exemption, unlike OpenAI and Perplexity. The docs don't state it in the words 'honors robots.txt', so that reading is an inference.
- Which IP list should I use to verify MistralAI-User?
- Use mistralai-user-ips.json. MistralAI-Index has its own list (mistralai-index-ips.json), and the docs give no IP list for MistralAI-Training.
Related crawlers
Sources
Checked against these pages on 2026-10-18. Operators change their crawlers; check the source before relying on a detail.