In short. Mistral Work's Open URL reads the supplied page and does not follow its links or crawl the rest of the site.
Definition: Mistral Work's Open URL feature fetches and reads the page at a URL supplied in a chat. Its documented retrieval boundary is one page: it does not crawl the rest of the site or follow links from that page.
Key facts
- Mistral describes Open URL as a way to analyze a page whose link is already available.
- Web Search must be enabled before Work can browse, and the documented activation path is the
+icon or/, then Tools, then Web Search. - A URL can be pasted with a question or instruction. Work fetches the page and uses its content as context, then shows a link icon and an
Opened Pagemention. - Online PDFs and documents are supported as URL targets, according to the same Work documentation.
- Several URLs can be pasted in one chat, and Work keeps the content from each page in context for comparison or follow-up questions.
- Mistral's crawler documentation assigns
MistralAI-Userto user actions in Vibe andMistralAI-Indexto automated web crawling for search indexing.
What does Open URL read?
Open URL is a page retrieval feature. Mistral says that a URL can be pasted into the message box with a question or instruction, after Web Search is enabled. The documented sequence ends with Work fetching that page and using its content to answer. The response includes an Opened Page mention, which Mistral presents as confirmation that the page content was read.
The input can also be an online PDF or document. For several supplied URLs, Work keeps the content from each page in context, and Mistral gives a comparison between two competitor product pages as an example. Each page in that context comes from an explicitly supplied URL.
Does it follow links to child pages?
Mistral states the boundary directly: “Open URL fetches only the single page at the URL you provide. It doesn't crawl the entire site or follow links.” A link from an opened page is therefore a reference visible in that page, rather than an instruction for Open URL to retrieve the linked page.
The same documentation lists three access conditions that affect the supplied page itself. Pages behind a login or paywall cannot be accessed. Some highly interactive sites may load incompletely. Web search can also return outdated or incomplete content depending on what the source page exposes. These are page-access limits, separate from the single-page retrieval boundary.
Which Mistral crawler handles a user-directed visit?
Mistral's crawler documentation says that MistralAI-User is for user actions in Vibe and that Vibe may visit a web page to help answer a question and include a source link. The page also gives this user-agent string:
Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; MistralAI-User/1.0; +https://docs.mistral.ai/robots)
The same page gives MistralAI-Index a different role: automated crawling for indexing, with indexed content used for Mistral search that helps answer questions in Vibe. The documented crawler roles describe why a Mistral request may appear in server logs; they do not change Open URL's one-page behavior.
Related terms
FAQ
Can I give Open URL several pages to compare?
Yes. Paste several URLs in one chat with the comparison request. Mistral says Work keeps the content from each supplied page in context.
Will a link on an opened page trigger another fetch?
No. The documented scope is the single page at the supplied URL, with no site crawl or link following. A linked page needs its own URL in the chat.
How can I tell that Work read the page?
Look for the link icon and the Opened Page mention in the response. Mistral documents these as indicators that Work read the page content.
Sources
Check how a page reads to answer engines. Open the GEO score analyzer