Skip to main content
Lesson 3 of 16 · Mistral Vibe: AI Coding CLI

First Launch of Vibe

Lesson 3 of 16 Updated November 2, 2026 Developer
On this page
  1. Launching Vibe for the First Time
  2. The Initial Setup Wizard
  3. The Interactive Interface
  4. Your First Exchange
  5. Launching with a Direct Prompt
  6. Returning to Configuration
  7. Key Takeaways

Lesson updated November 2, 2026 based on Vibe Code documentation: browser-based login is now the default, keys are saved in the system keychain, and the built-in model is Mistral Medium 3.5 (Devstral 2 was removed from the API on July 31, 2026). Sources: Install and setup · Configuration · Agents.

November 2, 2026 Update — Devstral. Devstral 2 (devstral-2512) reached its sunset date on July 31, 2026, and Devstral Small 2 on March 31, 2026: calls to these models are no longer guaranteed. The official table redirects them to Mistral Medium 3.5, which is itself deprecated in favor of Mistral Large 4. Codestral (codestral-2508) remains generally available for code completion. Check the model table to see where -latest aliases point before reusing these examples.

Launching Vibe for the First Time

Vibe always works within the context of a directory. Navigate to an existing project, not your home folder, before launching it:

cd /path/to/your/project
vibe

On the first launch, Vibe performs several automatic initialization steps. Don’t panic if the process takes a few seconds: the tool is creating its configuration directory and preparing the setup wizard.

The Initial Setup Wizard

The first time you run Vibe, an interactive wizard guides you through the configuration. It starts with the visual theme of the terminal interface: several color palettes are available. Choose the one that is most readable on your terminal. Nothing is set in stone; you can change this at any time.

Next comes the crucial step: accessing the models. The wizard offers two paths:

  1. Launch browser: You log in with your Mistral account in the browser, and Vibe uses your Mistral plan (Vibe Code is included, with usage limits depending on your plan and pay-as-you-go settings). This is the default path, requiring no key management.
  2. Use an API key: You generate a key from Code › Vibe CLI, or use a key from another compatible provider, and paste it into Vibe.

Vibe saves the key in your operating system’s keychain (macOS Keychain, Windows Credential Manager…) or, failing that, in ~/.vibe/.env, which is readable only by you. You can relaunch this wizard at any time with vibe --setup.

You won’t be asked to select a model: Vibe ships with a built-in model, Mistral Medium 3.5 (alias mistral-medium-3.5), which its documentation recommends for reasoning and coding. The /model command will allow you to choose a different one later from those you have configured.

The Interactive Interface

Once configuration is complete, you are in interactive mode. The screen is organized simply: an input area at the bottom for typing messages, the conversation history above, and indicators showing the model in use and session status. There is nothing else to learn to start working.

However, a few keyboard shortcuts will make a huge difference in your workflow speed. Memorize them now rather than looking them up a month from now:

ShortcutEffect
Ctrl+J or Shift+EnterNew line in your message (for multi-line input)
@File autocompletion (e.g., @src/main.py)
!Prefix to execute a shell command directly
Ctrl+GOpen an external editor to draft a long message
Ctrl+OShow or hide tool output
Cmd+\ or Alt+\Show or hide the task list (todo)
Shift+TabToggle between approval modes

The most useful shortcut for daily work is undoubtedly @: instead of describing where a file is located, you reference it directly, and the agent reads it.

Your First Exchange

Test it with a simple request. In a Python project, for example, type:

Explain the structure of this project

Watch what happens, because this is where the difference from a standard chatbot becomes visible. Vibe scans the project tree—files, folders, configuration—reads the Git status to identify the branch, recent commits, and modified files, analyzes key files like the README, package.json, or pyproject.toml, and then provides you with a structured summary of the architecture.

In other words, it doesn’t just answer; it fetches the information it needs. Ask the same question in an empty repository, and the response will be sparse—there is nothing to read.

Launching with a Direct Prompt

Interactive mode isn’t mandatory. For a single request, pass your query directly as an argument:

vibe "Refactor the main function to use async/await"

For programmatic mode—the kind you would use in a script or a continuous integration pipeline—the --prompt option combined with a usable output format works better:

vibe --prompt "Analyze test coverage" --output json

Two safeguards are particularly useful in this context, as an unsupervised agent can loop or become expensive. --max-turns limits the number of back-and-forth exchanges, and --max-price caps the spending:

vibe --prompt "Fix bugs" --max-turns 10 --max-price 0.50

Returning to Configuration

Sooner or later, you will want to change the theme or reconfigure your API key. The setup wizard can be relaunched on demand:

vibe --setup

During a session, slash commands take over: type / to see them appear. /help or /h displays full help, /config opens the configuration, /theme changes the theme, /model switches the model without leaving the session, and /clear erases the conversation history when you move to a completely different topic.

Key Takeaways

  • Launch Vibe with vibe inside your project directory
  • On the first launch, the wizard configures the theme and API key
  • The API key is saved in ~/.vibe/.env
  • Mistral Medium 3.5 is the built-in model; use /model to switch models
  • Keyboard shortcuts (@, !, Ctrl+J, Shift+Tab) are essential for daily use
  • Use vibe "your question" for a quick exchange without interactive mode