Mistral the Cat turned Vibe: the guide to using it in business (2026)
This guide was published on October 24, 2026, and is based on Mistral AI's official announcement, help center, documentation, and pricing page, which were consulted on that day. Mistral updates Vibe monthly; check your account for details.
Mistral Chat no longer exists under that name: since May 28, 2026, Mistral AI's assistant is called Vibe. The address remains the same (chat.mistral.ai), accounts, conversations, and subscriptions have been preserved, but the product is no longer just a chatbot. It's an agent that schedules and executes tasks within your tools (Work mode), combined with a development agent (Code mode), all under a single license. This guide reviews what has changed, what Vibe can do, what it costs, what it does with your data, and how to properly introduce it into a business.
The essentials in five lines
- The Cat has become Vibe May 28, 2026: same address, same login details, history retained.
- Three modes: Work to delegate tasks in several stages in your tools, Code to develop up to the merge request, Chat for classic conversation, destined to disappear in favor of Work.
- Price: Free with limits, Pro at €14,99 excl. VAT per month, Team at €24,99 excl. VAT per user per month, Enterprise on quote.
- Data: servers in Europe; in free and Pro, your exchanges are used for training as long as you do not disable it; in Enterprise, they are excluded by default.
- Distinctive feature: a French publisher, open-weight models and the possibility of deployment in your private cloud or on your servers.
From Le Chat to Vibe: what changes, what remains the same
Mistral presented Vibe as "an agent and a license" for work and code. The name isn't new: it was already the name of Mistral's command-line development tool, which was merged with the former Le Chat. For a user, the switchover was seamless, without any migration.
| The Cat (until May 2026) | Vibe (since May 28, 2026) | |
|---|---|---|
| Address | chat.mistral.ai | chat.mistral.ai, and chat.mistral.ai/code for Code mode |
| Main use | Conversation, research, writing, document analysis | Delegation of tasks in several stages, with a plan to be validated. |
| Code | Separate tool (Mistral Vibe command-line tool) | Integrated: web, VS Code extension and command line, same license |
| Connectors | Connectors and MCPs added in 2025 | Over 100 connectors, active in Work mode |
| Accounts and history | Mistral account, conversations, libraries | Preserved exactly as they were |
| Mobile Apps | The Chat | Vibe, on the App Store and Google Play |
Don't confuse Vibe, Mistral's product, with "vibe coding," the practice of having AI write an application by describing it in natural language. Vibe's Code mode allows for vibe coding, but the term refers to a method, not a tool: we detailed it in Vibe coding in the workplace: where to allow it, how to regulate it.
The three Vibe modes
Vibe Work: Delegate an entire task
The new feature justifies the name change. You describe an expected result; Vibe proposes a plan, awaits your approval, and then executes it through your connected tools. Each step remains visible, and you can expand the tool calls to see what the agent has read and produced. Mistral highlights five use cases:
- searching within the company's knowledge base (documents, email, shared spaces);
- Data analysis: connect a database or upload a spreadsheet to highlight trends and anomalies, with graphs and dashboards in the conversation;
- the synthesis of documents and reports in the Canvas tool (notes, reports, responses to calls for tenders, committee materials), with sending the deliverable to Notion, SharePoint or your email inbox;
- scheduled tasks, executed once or daily, weekly or monthly, with a notification at the end of each execution;
- the skillsreusable procedures, ready-to-use or created by you, based on open standards.
Vibe Code: Developing with an agent
Code mode connects to GitHub and guides a development request through to the merge request (Pull requestThe agents work in isolated environments, in parallel, and continue even when your computer is turned off; you approve sensitive actions and review changes as they are made. It is available as a web application, a VS Code extension, and a command-line application, with the ability to transfer a session from the terminal to the cloud.
Vibe Chat: the classic conversation
The question-and-answer functionality of the old Le Chat remains available. However, Mistral has announced that it will soon be discontinued in favor of the Work mode, which retains the history. In practice, Work also answers simple questions; the difference lies in its ability to chain actions together.
What Vibe can do on a daily basis
| Function | What is it used for? |
|---|---|
| Web search and in-depth research | Sourced answers and multi-stage research reports on a topic |
| Document libraries | Grouping files (procedures, contracts, studies) that Vibe consults to respond |
| Canvas | Writing and revising a long document alongside the conversation |
| Spreadsheets and presentations | Generate Excel files and slides (paid options) |
| Connectors | Google Workspace, Outlook, SharePoint, Slack, GitHub, Jira, and your own MCP connectors |
| Images and voices | Image generation and editing, dictation and voice chat |
| Memo | Remember your preferences and the context from one conversation to another |
| Agents and skills | To encapsulate a repeated procedure and share it with the team |
Under the hood, Vibe relies on Mistral's flagship models, in particular Mistral Medium 3.5, an open-weight model whose level of reasoning can be adjusted: quick response for reformulation, longer reflection for analysis.
Offers and prices in October 2026
| It offers | Price excluding taxes | For who | What it brings |
|---|---|---|---|
| Free | Free | Personal discovery | Limited messaging, searches and code sessions, images, over 100 connectors |
| Pro | 14,99 € per month | Intensive individual use | Limits raised, faster responses, all-day coding, spreadsheets and presentations, API credits; discounted pricing for verified students |
| Team | €24,99 per user per month (€19,99 with annual billing) | Teams and SMEs | Shared space, up to 30 GB of storage per user, domain verification, data export, administration |
| Enterprise | Quote only | Large organizations | Custom templates, agents and workflows, SAML SSO, audit logs, white labeling, private cloud or on-premises deployment, dedicated support |
Pro and Team have the same usage limitations: what Team adds is the shared workspace and administrative tools. For a company, both the free and Pro versions present the same difficulty: they are individual accounts that the organization cannot manage.
Your data: training, hosting, sovereignty
This is the point to be addressed before any deployment, because the default setting depends on the offer.
| It offers | Are your exchanges used for training? | Who can disable it? |
|---|---|---|
| Free and Pro | Yes, by default | The user, in the settings (on mobile: Settings, then Data & Account Controls, uncheck "Enable data sharing") |
| Team | According to the organization's settings | The administrator, for everyone: Admin panel, Manage, Vibe, Privacy, option “Allow your interactions to be used to train our models” |
| Enterprise | No, by default | The administrator, if he chooses to activate it |
| API (Mistral Studio) | Separate setting | Admin panel, Privacy, “Anonymous improvement data” |
Two pitfalls to be aware of. Vibe and API settings are independent: disabling one does not disable the other. And a thumbs up or down accompanied by a comment authorizes Mistral to use the relevant exchange to improve its models, regardless of the offer.
Regarding hosting, Mistral states that its servers are located in Europe, specifically in Sweden. As the company is incorporated under French law, the issue of data sovereignty often influences the choices of public organizations and regulated sectors seeking to limit their exposure to extraterritorial US laws. The Enterprise offering goes further: deployment in your private cloud or your own data centers, where data remains within your environment. Open-weight models can also be hosted by you, under a Mistral commercial license for enterprise use.
Vibe versus Copilot, ChatGPT and Claude
| Vibe (Mistral AI) | Microsoft 365 Copilot | ChatGPT (OpenAI) | Claude (Anthropic) | |
|---|---|---|---|---|
| Editor | French | American | American | American |
| strong point | Sovereignty, open models, on-site deployment | Native integration with Outlook, Teams, Word and Excel | Versatility, GPT and agent ecosystem | Drafting and analyzing long documents, delegation with Cowork |
| Delegation of tasks | Work Mode | Copilot Cowork | Agent mode | Claude Cowork |
| Code agent | Vibe Code | GitHub Copilot | Codex | Claude Code |
| Hosting on your servers | Yes, in Enterprise | No | No | No |
The right choice depends less on performance, which is now quite similar, than on your environment: an organization using Microsoft 365 will primarily benefit from Copilot, while an organization handling sensitive data or operating in the public sector will first consider Vibe. Many organizations combine the two, with Vibe used for the most sensitive files. For more information, see our comparison. Claude or ChatGPT and the guide Which AI tool should you choose for your profession?
Five missions to entrust to Vibe Work to begin
These examples assume the corresponding connectors are enabled. Always specify the expected result, the source to use, and the output format, then review the plan before validating it.
- Preparing for a meeting. "Based on the email exchanges of the last two weeks with client X and the latest report in SharePoint, prepare a one-page memo: open points, expected decisions, questions to ask."
- Analyze a spreadsheet. "Here is the export of sales figures for the quarter. Compare each region to the previous quarter, highlight the three largest differences, and provide an explanation for each to be verified. Add a graph."
- Responding to a call for tenders. "Read the attached specifications and our reference library. List the requirements, indicate for each the reference that best meets it, and mark those for which we have nothing."
- To conduct regular monitoring. "Every Monday at 8 a.m., research the week's news about our three main competitors and send me a ten-line summary with the links."
- Sorting an email inbox. "Go through my unread messages since Friday, categorize them as 'to deal with today,' 'to delegate,' 'for information,' and provide a short reply for the first five."
Deploying Vibe in a company: six steps
- Choose the offer. A Team account is recommended for use by more than two people for work; an Enterprise account is recommended if you need SSO, audit logs, or on-premises hosting. Avoid allowing teams to work on free personal accounts.
- Adjust privacy settings. Check the training setting in the administration, and the equivalent option on the API side if your technical teams use it.
- Establish the rules of use. What data can be entrusted to Vibe, which never, who reviews what: a short charter is sufficient, the AI charter template provides a basis for adaptation.
- Open the connectors gradually. Start with documents and email for a pilot service, under minimal privileges, before expanding. An agent who reads everything can also summarize everything for anyone.
- Train using real-life cases. Delegation cannot be improvised: it involves formulating a mission, reviewing a plan, and verifying results. This is the purpose of our... Mistral training in companiesfrom the first prompt in Vibe to scheduled tasks, on your team's files.
- Measure and adjust. Number of active users, recurring missions implemented, time saved declared: three indicators are enough to decide whether or not to expand.
A Vibe deployment also counts towards the AI Act: training users is one of the AI literacy measures expected since February 2025. Our article AI Act: What your company needs to do details the schedule and obligations.
F.A.Q
Does Mistral's Cat still exist?
Not under that name. Since May 28, 2026, Le Chat has been called Vibe. The address chat.mistral.ai, usernames, conversations, and subscriptions have remained the same.
Is Vibe de Mistral free?
Yes, a free plan exists, with limits on messages, searches, and coding sessions that Mistral doesn't specify precisely. The Pro plan costs €14,99 excluding VAT per month, and the Team plan costs €24,99 excluding VAT per user per month.
Where is my data stored?
Vibe is deployed on Mistral AI's servers, located in Europe (Sweden), according to the publisher. With the Enterprise offering, Vibe can be deployed in your private cloud or on your own servers.
Is Mistral using my conversations to train its models?
In the free and Pro plans, yes, it's enabled by default, and you can disable it in the settings. In the Team plan, the administrator sets the option for the entire organization. In the Enterprise plan, data is excluded from training by default.
What is the difference between Vibe Work and Vibe Code?
Work delegates office tasks through your tools (email, documents, spreadsheets, calendar). Code delegates software development to your GitHub repositories, up to and including the merge request. Both are included in the same license.
Is Vibe compliant with the GDPR?
Mistral offers a data processing agreement and hosts its servers within the European Union, which simplifies compliance. However, compliance depends primarily on your usage: legal basis, informing individuals, data entrusted to the tool, and training settings. These points fall under the purview of your Data Protection Officer (DPO).
Sources
- Mistral AI, announcement from Vibe, the unified agent for work and code, May 28, 2026
- Mistral AI Help Center, The Chat is now Vibe.
- Mistral AI, Vibe offers and prices, accessed October 24, 2026.
- Mistral AI Help Center, disable the use of your data for training et use of user data.
- Mistral AI Help Center, data localization.
- Mistral AI documentation, Vibe presentation.
- The Decoder, Mistral renames Le Chat en Vibe, May 28, 2026