Configuration - Mistral AI

2026-05-28 · Source: mistral.ai via Google News · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Software Development & Engineering · Depth: Intermediate, short

Summary

The Vibe Code CLI is configured through a `config.toml` file, found at project-level (`./.vibe/config.toml`) or user-level (`~/.vibe/config.toml`), with project settings overriding user settings. This file is accessible via OS-specific commands or the `/config` slash command. The CLI's default configuration and state directory, `~/.vibe/`, can be customized using the `VIBE_HOME` environment variable, which impacts `config.toml`, `.env`, and related directories for agents, prompts, skills, tools, and logs. Command-line flags like `--workdir` also define project context. Key `config.toml` sections allow setting a `default_agent`, defining `providers` and `models` (e.g., `active_model = "devstral-openrouter"`), configuring `MCP servers`, managing `skills` (e.g., `enabled_skills = ["code-review", "test-*"]`), and setting `tool permissions and filters` (e.g., `enabled_tools = ["serena_*"]`). Update and telemetry settings, including `enable_auto_update`, `enable_notifications`, and `enable_telemetry` (all defaulting to `true`), are also configurable. Configuration precedence prioritizes command-line flags, then environment variables, followed by project-level, and finally user-level `config.toml`.

Key takeaway

For AI Engineers managing Vibe Code CLI deployments, understanding its configuration hierarchy is crucial for predictable behavior. You should explicitly define `config.toml` settings at the project level for specific project requirements, ensuring they override user-level defaults. Leverage `VIBE_HOME` to standardize your Vibe environment across projects and use command-line flags for session-specific overrides. Regularly review `enabled_skills` and `tool permissions` to maintain agent control and security, and consider disabling telemetry if an offline-only setup is required.

Key insights

Vibe Code CLI configuration is managed through a hierarchical `config.toml` system, environment variables, and command-line flags.

Principles

• Project config overrides user config.
• CLI flags have highest precedence.
• `VIBE_HOME` centralizes Vibe data.

Method

Configure Vibe Code CLI by editing `config.toml` at user or project level, setting `VIBE_HOME`, or using command-line flags to define agents, models, skills, and tool permissions.

In practice

• Set `active_model` for specific LLMs.
• Use `enabled_skills` to control agent capabilities.
• Disable telemetry for offline setups.

Topics

• Vibe Code CLI
• Configuration Management
• TOML
• Environment Variables
• Agent Skills
• Tool Permissions
• Model Context Protocol

Best for: AI Engineer, Machine Learning Engineer, MLOps Engineer

Related on AIssential

• Two users, one CLI: people and agents - Mistral AI — Designing CLIs for AI agents improves tool composability, scriptability, and testability for all users.
• The End of Vibe Coding — Ambiguous instructions, or "vibe coding," lead to AI agent failures; precise specifications are key.
• Mistral Vibe - Mistral AI — Mistral Vibe enables terminal-based, agentic code generation with contextual awareness and user-controlled guardrails.
• Vibecoder final boss — AI agents with filesystem access pose a real security risk for exposing sensitive `.env` files.
• Vibe DevOps is Already Happening? — Vibe DevOps involves direct SSH access to production for deployments and maintenance, eschewing formal infrastructure-as-code.
• QCon London 2026: Morgan Stanley Rethinks Its API Program for the MCP Era — Codifying API architectures with CALM enables scalable, secure, and efficient integration for AI agents in financial institutions.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by mistral.ai via Google News.