Platform Overview

The core conversational interface powered by Mistral's latest models. Features multi-turn context retention, multilingual support, and a responsive multi-modal experience for everyday productivity tasks.

Le Chat autonomously plans a research strategy, navigates the web to synthesize multiple trusted sources, and generates long-form analytical reports.

Upload PDFs, spreadsheets, and Word documents directly into chat. Le Chat applies OCR 3 natively to extract text from scanned images and complex document layouts.

A secure Python sandbox directly within Le Chat. Upload datasets to explore data, generate charts, and automate data transformations without leaving your browser.

An integrated writing and coding surface. Le Chat generates text, code, or presentations, allowing you to collaboratively refine and edit the AI's output in real time.

Enable Memories to let Le Chat retain your preferences across sessions. Build Libraries from uploaded documents to ground the model's responses in your own secure knowledge base.

Create AI assistants for specific workflows. Define custom system instructions, attach tools (like MCP connectors), link knowledge bases, and share them securely with your team.

A visual lab to experiment with system prompts, user messages, and generation parameters (like Temperature and Top-P). Test prompts side-by-side across different Mistral models instantly.

Generate and securely manage programmatic access tokens. API keys are scope-limited to specific workspaces, allowing precise control over which applications can access specific resources.

Train custom Mistral models on your own datasets via JSONL file uploads. Adapt base models to your domain terminology, tone, or specific formatting tasks.

Run benchmark pipelines against your fine-tuned deployments or complex system prompts to objectively measure accuracy, performance, and detect regressions over time.

Deploy custom agents into production. The Agents API allows developers to programmatically execute workflows where Mistral models autonomously orchestrate external tool calling.

A conversational interface to your codebase, right in the terminal. Reference files with @, run shell commands with !, and use slash-command skills to automate common workflows.

Build custom agents with system prompts and reusable slash-command skills. Automate repetitive workflows like code review, test generation, and documentation updates.

Integrate Mistral Vibe with JetBrains and Zed via the Agent Communication Protocol. Get AI assistance directly in your editor while keeping terminal access.

Run Mistral Vibe offline with locally deployed models. Keep your code on your machine with no data leaving your environment, while retaining full agent capabilities.

Centrally administer your organization. Invite team members, enforce secure Admin/Billing/Member roles, and allocate paid seats for premium products like Vibe Team and Mistral Code.

Mandate corporate security policies by connecting Mistral Platform to your existing identity provider via SAML SSO. Includes secure domain name verification to claim organizational domains.

Divide your organization into distinct, secure workspaces. Isolate API keys, fine-tuned models, and billing groups so that different departments operations never overlap.

Monitor token expenditure in real time. Download invoices, update payment methods, and establish strict hard spending caps per workspace to completely prevent unexpected cost overruns.

Maintain a strict, chronological record of organizational events. Track when users are invited, when API keys are generated, and view security events for internal compliance auditing.

Take full control of data sovereignty. Admins can globally opt out of data collection for model training, manage data retention policies, and execute strict GDPR deletion requests.