Enterprise Search

Updated November 2, 2026 — Le Chat is now Vibe. As of May 28, 2026, Le Chat has been rebranded as Vibe (still at chat.mistral.ai): a single agent for work and code that preserves your conversations, settings, and subscriptions. The legacy Le Chat experience remains available in Chat mode; in Work mode, agents are now skills (invoked with @), memories form the knowledge base, and deep research is the Deep Research skill. Plans are now called Mistral Free, Pro, Team, and Enterprise. This lesson uses the name Vibe; note that some menus described here may have changed names or locations.

Knowledge is inherently scattered

Organizational knowledge never lives in one place. It spreads across cloud storage, collaboration tool threads, internal knowledge bases, wikis, CRMs, and support tickets. Every employee knows roughly where to look within their own area, but almost no one knows how to search outside it. This fragmentation costs you: repeated questions, duplicated work, and decisions based on outdated document versions.

Enterprise Search solves this with unified, AI-powered search built directly into Mistral Enterprise. Instead of opening four different tools in sequence, an employee asks a question in natural language and receives a synthesized answer drawn from connected sources. The difference shows up on mundane questions: “What is the travel expense reimbursement policy?” stops being a hunt across three document spaces and becomes a single sentence with a sourced answer.

Understand intent, not just keywords

Traditional internal search engines rely on keyword matching, and that is their limit. Search for “new employee onboarding procedure” when the document is titled “Recruit Integration Guide”: classic search returns nothing, not because the information is missing, but because the vocabulary differs. Enterprise Search relies on the semantic understanding of Mistral models to interpret the intent behind the question and retrieve relevant documents despite phrasing gaps, recognizing synonyms, alternative phrasings, and your organization’s specific business context.

This understanding rests on a four-step indexing process. First, Enterprise Search connects to your tools via dedicated connectors, then extracts and normalizes the documents, messages, and data they contain. Next, it transforms each piece of content into a semantic vector using Mistral’s embedding models—a numerical representation of meaning rather than words. Finally, indexing is incremental: only new or modified content is processed, so a report published late morning becomes queryable immediately without reprocessing the entire corpus.

Available connectors

Coverage determines utility, and native connectors target the platforms most common in enterprises. For document storage, Google Drive brings documents, spreadsheets, presentations, and shared files; Microsoft SharePoint contributes sites, document libraries, and lists; OneDrive handles personal and shared files; Dropbox Business covers team folders; and Notion provides pages, databases, and team wikis.

On the communication side, the Slack connector indexes public messages and authorized channels; Microsoft Teams covers channel conversations and shared files; Confluence handles spaces, pages, and blog posts. For technical teams, GitHub exposes issues, pull requests, documentation, and wikis; GitLab offers issues, merge requests, and wikis; Jira provides tickets, epics, and sprints—enough to uncover the rationale behind an architecture decision buried in a ticket no one can name anymore. On the sales side, Salesforce supplies accounts, contacts, opportunities, and knowledge articles; Zendesk delivers tickets and help center articles; HubSpot provides contacts, companies, and deals.

How a query travels

A search passes through three stages. First, the model analyzes the question: it extracts intent, identifies key entities, and optionally reformulates the query to maximize relevance. Next comes the actual search, which combines two approaches. Semantic search compares the query vector against indexed document vectors to surface content close in meaning; lexical search, the classic keyword approach, ensures coverage of terms no model can guess—internal project names, product codes, homegrown acronyms. This hybrid approach lets you find a document whose title bears no resemblance to your question while still catching the one that contains the exact reference “PRJ-2231.”

The final stage synthesizes found information into a clear answer, complete with precise citations linking back to source documents. Every claim remains traceable to its origin, letting you verify before acting and dig deeper if needed. Build this habit early in your teams: open at least one source before incorporating an answer into a deliverable. That simple step separates professional use from blind trust.

Users see only what they are allowed to see

Permission handling often decides whether security leadership approves the service. Enterprise Search never grants access to a document a user could not otherwise consult: rights are inherited from source systems—Google Drive, SharePoint, or any other connected tool—and results are filtered accordingly for each person. These permissions sync regularly with origin systems, so a rights change—an employee switching teams, a folder becoming restricted—reflects immediately in results. The tool creates no backdoor in your authorization model; it depends on it entirely. All the more reason to clean up overly broad shares before going live: Enterprise Search will make queryable everything your users had the right to read without ever suspecting its existence.

Key takeaways

• Enterprise Search delivers unified natural-language search across all internal data
• Semantic indexing goes beyond keywords to understand query intent
• Native connectors cover major platforms (Google, Microsoft, Slack, Confluence, GitHub, and more)
• Hybrid search combines semantic and lexical methods for optimal coverage
• Existing permissions are respected: each user sees only what they are authorized to access