
Vibe
Mistral's open-source CLI coding agent, powered by Devstral, for terminal-based agentic workflows.
Last refreshed: 17 May 2026 · Appears in 1 active topic
Is Mistral Vibe the first serious EU-built alternative to Claude Code for enterprise developers?
Timeline for Vibe
Bundled as coding agent in Le Chat Enterprise stack
European Tech Sovereignty: Mistral ships Le Chat Enterprise and Medium 3.5What is Mistral Vibe and how does it compare to Claude Code?
How do I install Mistral Vibe CLI on my machine?
Does Mistral Vibe work inside VS Code or only in the terminal?
Background
Mistral Vibe is an open-source command-line coding agent launched by Mistral AI in December 2025 alongside Devstral 2. It connects to Devstral models and enables end-to-end autonomous code modification — reading files, writing code, running tests, and committing changes — entirely through natural language instructions in the terminal. Vibe was introduced as Mistral's entry into the competitive agentic-coding market dominated by tools such as Claude Code and GitHub Copilot Workspace. The launch formed part of Mistral's broader Le Chat Enterprise announcement .
Vibe is released under the Apache 2.0 licence and integrates via the Agent Communication Protocol (ACP), allowing it to run as a native extension in IDEs such as Zed in addition to the terminal. In April 2026 Mistral shipped Vibe 2.0, switching the default backbone model from Devstral 2 to Medium 3.5 and adding remote cloud-agent mode — the agent executes in Mistral's infrastructure rather than consuming local compute, handling complex multi-file rewrites asynchronously. Mistral partnered with Zed IDE to embed Vibe as a native extension.
Vibe's significance lies in providing a fully EU-origin, self-hostable alternative to Claude Code (Anthropic, US) and GitHub Copilot Workspace (Microsoft, US). For European enterprises and public bodies subject to data-residency mandates, an open-source terminal agent that can run entirely on Hetzner or OVHcloud infrastructure — with no US API calls — closes a gap that no other EU AI lab has addressed at the agentic layer.