Mistral Vibe, Formerly Le Chat: Complete Guide to Features and Workflows

by David Porter

Friday, 28 August 2026 at 07:11

thumbnail_what-is-mistral-ai-models-comp

Mistral Vibe is Mistral AI’s unified agent for professional work and software development. It combines the conversation history and familiar features of Le Chat with a newer Work environment for multi-step tasks and a Code environment for repository-level development.

Mistral announced the change on 28 May 2026. Existing Le Chat accounts, plans, conversations and settings carried over, while chat.mistral.ai remained the main entry point.

The product now has three distinct modes:

Mode	Best for	Where it runs
Work	Research, documents, connected apps and scheduled tasks	Web and mobile
Chat	Quick conversations and legacy Le Chat features	Web and mobile
Code	Reading repositories, editing code, running commands and opening pull requests	Web, terminal and IDE

For the wider company and model story, start with What Is Mistral AI?. This guide focuses on using Vibe as a product.

What happened to Le Chat?

Le Chat has become part of Vibe rather than disappearing overnight.

Mistral previously used the Vibe name for its terminal coding assistant, introduced as Mistral Vibe CLI in December 2025. The May 2026 change expanded that brand across the company’s work and coding products. Le Chat users were moved into the same Vibe account and subscription structure.

The current interface still includes a Chat tab. Mistral describes it as the home for quick turn-based conversations and several legacy Le Chat features. Work handles the newer agentic workflows, while Code covers software engineering.

That distinction matters when following older tutorials. Instructions that say “open Le Chat” may still lead to the correct website, but the relevant feature could now sit under Work, Chat or Code.

Our Mistral hub tracks the product as its naming and capabilities change.

Vibe Work: delegate a complete task

Work is designed for requests that require several steps, sources or tools.

A user describes the intended result. Vibe can gather context, create a plan, search approved sources, call connected tools and produce a deliverable. The interface shows its progress and tool calls while it works.

According to Mistral’s current Work documentation, available context can include:

uploaded documents, spreadsheets, presentations, PDFs and images;
Libraries containing indexed source collections;
public web pages and web search;
email and calendar services;
Slack, Notion, GitHub, Google Drive and SharePoint;
reusable Skills and custom instructions;
Projects that group related work;
and Canvas for editable documents, presentations, code or mock-ups.

Useful Work requests include:

Research these five competitors using their official documentation and create a one-page comparison for our sales team.

Review this contract, identify renewal dates and liabilities, and show the supporting clause for every finding.

Find this week’s customer feedback in Notion and Slack, group it by theme and draft a product update.

The quality of the result still depends on access, source quality and instructions. Give Vibe the desired outcome, audience, source boundary, format and acceptance criteria. Review the finished document against the original material.

Scheduled work

Vibe can run prompts once or on a daily, weekly, monthly or yearly schedule. Mistral currently lists up to five scheduled tasks on Free and unlimited scheduling on Pro and Team, subject to fair-use limits. Give recurring agents a narrow job, restricted access and a clear point at which they must ask for human input.

Vibe Chat: quick answers and older Le Chat tools

Chat preserves the shorter prompt-and-response experience. It is the simplest mode for questions, drafting, translation, brainstorming and other tasks that do not require a long agent run.

It also contains several features inherited from Le Chat:

Agents: configured assistants with their own instructions and tools;
Think mode: extended reasoning for more difficult questions;
Code Interpreter: sandboxed code execution and data analysis;
Memories: saved user facts and preferences;
Deep Research: now also available in Work as a Skill.

Mistral’s mode comparison describes these as legacy Chat features that are being reworked for Work. There is no reliable public date for the complete retirement of the Chat tab, so a guide should describe the present interface rather than promise a migration deadline.

Use Chat when you want an immediate exchange. Use Work when the result depends on several sources, actions or checkpoints.

Vibe Code: terminal, IDE and remote agents

Code gives Vibe access to a software project under user supervision. Depending on the chosen surface, it can inspect files, edit code, run shell commands, execute tests and prepare a branch or pull request.

The current options include:

Vibe CLI: a terminal-native coding agent for a local checkout;
IDE integration: Mistral lists support for VS Code and JetBrains environments;
Vibe Code Web: remote sessions running against connected GitHub repositories in managed sandboxes.

Remote sessions can continue while the user’s computer is offline. Several sessions can work in parallel, and Teleport can move an active session between local and remote environments.

A productive coding request should identify the repository, desired behavior, architecture constraints, prohibited files and required test commands. For example:

Add CSV export to the invoices page. Preserve the existing API, follow the current test patterns, run the relevant unit and integration tests, and show me the diff before creating a pull request.

Vibe can generate a large amount of plausible code quickly. Compilation, tests, security checks and human review remain necessary. A successful agent run means the change survived the project’s acceptance process, not that the model produced a confident summary.

Approvals and tool control

Work asks for approval before sensitive actions such as sending email, posting a message, changing an issue, creating an event or deleting data. Code similarly exposes plans, commands and file changes.

Mistral’s safety documentation gives users three choices:

approve the specific action;
allow that function for the rest of the session;
decline and redirect the task.

“Always allow” removes a checkpoint. Keep broad writes, sends and deletes on manual approval until the workflow has been tested. Review read-only access too: a large searchable document collection may expose information the user did not expect the task to retrieve.

Vibe, the API and open-weight models are different products

Mistral Vibe is a hosted application. A Vibe subscription gives access to its interfaces, agent features and usage allowances.

The Mistral API is a developer service with separate credentials, model prices and pay-as-you-go billing. It can be used to build an application without using the Vibe interface.

Mistral also publishes models that can be downloaded or deployed through other infrastructure. The Mistral models guide explains which families are open-weight, commercially licensed or hosted-only. Our local Mistral guide covers hardware and runtimes.

These routes have different features and responsibilities. Running a Mistral model locally does not recreate Vibe’s connectors, remote agents, account history or support.

Which Vibe plan do you need?

Free is suitable for testing Chat, Work and limited coding. Pro raises the limits and is the practical individual plan for long-running tasks or regular development. Team adds a managed shared workspace, more storage and organizational controls. Enterprise covers private deployments, custom models, SSO, audit logs and negotiated support.

The complete rates, API credits and annual discounts are covered in our Mistral pricing guide.

Plan choice also changes data handling. Free and Pro users should check the model-training setting before uploading private material. Team and Enterprise inputs and outputs are excluded from model training under Mistral’s current policy. Read Is Mistral AI Safe? before connecting business systems.

Where Vibe fits

Vibe suits users who want one European platform for general work, scheduled agents and coding, with a route toward managed or private deployment. Google Workspace and Microsoft 365 users may still prefer the native reach of Gemini or Copilot, while ChatGPT and Claude have larger user ecosystems. Our ChatGPT alternatives comparison places them side by side. Test a complete workflow with non-sensitive material before moving a team.

Frequently asked questions

Is Le Chat now called Mistral Vibe?

Yes. Mistral announced in May 2026 that Le Chat had become Vibe. Existing accounts, plans, history and settings carried over.

Does the Chat mode still exist?

Yes. Vibe currently has a Chat tab for quick conversations and legacy features. Mistral is moving more capabilities toward Work, but it has not published a dependable final shutdown date for Chat.

What is the difference between Vibe Work and Chat?

Chat is designed for direct turn-by-turn conversations. Work plans and executes longer tasks across files, websites and connected tools, showing progress and asking for approval before sensitive actions.

Can Mistral Vibe write and run code?

Yes. Vibe Code can inspect files, edit code, run commands and tests, and prepare pull requests. It is available through the CLI, supported IDEs and remote web sessions.

Does a Vibe subscription include unlimited API use?

No. Vibe plans and API pay-as-you-go billing are separate. Free and Pro currently include limited monthly API credits, after which API usage follows the applicable model and tool rates.

Can I run Vibe completely on my own computer?

The Vibe CLI runs locally but can call hosted model APIs. Open-weight Mistral models can also be self-hosted, although that does not reproduce the complete hosted Vibe product.