Should you block MistralAI-User? Mistral's live-fetch agent
MistralAI-User is operated by Mistral. It fetches pages on demand when a user's request needs them, so what it reads feeds live answers and citations rather than a training set.
Who should block MistralAI-User, and who should not?
Leave it open for public pages: it visits because a Vibe user asked something, and Mistral says the answer can link to the source. Blocking it does not keep you out of Mistral's search index or training data, which belong to MistralAI-Index and MistralAI-Training. To opt out of Mistral training, disallow MistralAI-Training by name; to leave Mistral search, disallow MistralAI-Index.
The verified facts
| User-agent token | MistralAI-User/1.0 |
|---|---|
| Full user agent in logs | Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots) |
| Operator | Mistral |
| Purpose | Live retrieval and user-request fetchers |
| Honors robots.txt | Yes, per the operator's documentation. |
| Published IP ranges | mistral.ai/mistralai-user-ips.json |
What does Mistral’s documentation add?
- Mistral says MistralAI-User is for user actions in Vibe. It is not used to crawl the web automatically, nor to crawl content for generative AI training.
- Mistral documents two more agents with their own tokens: MistralAI-Index crawls automatically to index content for Mistral search, and MistralAI-Training builds datasets for training Mistral models.
- When it visits a page to help answer, Mistral says the response may include a link to the source.
How do you block MistralAI-User?
Add this to your robots.txt:
User-agent: MistralAI-User
Disallow: /Once you give a crawler its own group, it stops reading your User-agent: * rules, so run the whole file through the AI crawler robots.txt tester before you deploy it.
How do you tell a real MistralAI-User request from a fake one?
A user-agent match proves nothing, since any client can send MistralAI-User/1.0. Check the IP against the published range file above, re-fetched on a schedule. A request that fails is not Mistral. The log verification guide has a script for the IP check and a worked reverse-DNS example, or paste the IP into the AI crawler IP verifier, which runs these checks against Mistral’s current data.
What does blocking MistralAI-User cost you?
Vibe users cannot have Mistral's assistant visit your pages to answer a question.
What else does Mistral document?
Mistral also documents MistralAI-Index and MistralAI-Training, each with its own robots.txt token. How Mistral’s crawlers fit together.
Which crawlers in the same group should you decide on at the same time?
A robots.txt group for MistralAI-User does nothing to crawlers with other tokens. In the same group, the directory also covers ChatGPT-User (OpenAI), Claude-User (Anthropic), Perplexity-User (Perplexity), meta-externalfetcher (Meta) and Amzn-User (Amazon), each with its own token and its own documented cost of blocking.