
Mistral Vibe
Mistral's open-source agentic coding assistant — CLI, IDE extensions, and cloud remote agents.
About
Mistral Vibe is Mistral AI's open-source (Apache 2.0) agentic coding assistant, launched in December 2025. It runs as a terminal CLI, as native VS Code and JetBrains extensions, via the Zed editor through Agent Client Protocol, and as remote agents in the cloud. Powered by Devstral 2 and Mistral Medium 3.5 (provider-reported 77.6% on SWE-Bench Verified), it reads and edits files, executes shell commands, orchestrates subagents for parallel work, and extends through a skills system plus MCP server configuration. Sessions support plan-then-execute sign-off, session-scoped permissions, Git worktrees, and /teleport to move a live session between terminal and cloud without losing history.
The honest picture: Vibe is the newest entrant in a now-standardized category (skills, subagents, permissions, MCP) so it competes on model quality and execution rather than novel interface ideas. Hosted inference and higher usage limits require a Mistral plan — Free, Pro at $14.99/month, Team at $24.99/user/month, with API usage billed separately — and Free/Pro inputs may be used for model training unless you opt out. As a European-built, open-weight alternative to US-hosted agents it has a genuine differentiator, but the ecosystem (community skills, integrations) is still catching up to incumbents.
What it can do
- Terminal CLI coding agent with file editing and shell execution
- VS Code, JetBrains, and Zed editor extensions on one shared harness
- Remote cloud agents in isolated sandboxes with GitHub PR creation
- Subagents, reusable skills, MCP servers, and AGENTS.md project rules
- Tab-to-complete multi-line completions and voice mode
Setup
Install with curl -LsSf https://mistral.ai/vibe/install.sh | bash or uv tool install mistral-vibe. Sign in with a Mistral account, then run vibe in any project directory. IDE extensions install from the VS Code / JetBrains marketplaces.
Prerequisites
A Mistral account (free tier available); for hosted models, a Pro/Team plan or pay-as-you-go API key.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 58; Trending score 88/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.