Skip to main content

Configuring Gemini Permissions

The permission you grant with "Always" in the interactive prompt is currently session-specific, the permissions will be forgotten when you restart the CLI.

Here are the two primary ways to achieve persistent or automatic approval:

🚀 Option 1: Use YOLO Mode (--yolo)​

The easiest way to bypass all tool confirmations, is to start the Gemini CLI in YOLO mode. This setting is temporary for the session unless you use the second method.

  • To start with YOLO mode:

    gemini --yolo

    This mode will automatically approve all tool calls the model makes (like running shell commands, which includes npm), so use it with caution as it removes the security prompt.

  • To toggle YOLO mode on/off within a session: Press Ctrl+y inside the Gemini CLI prompt.

💾 Option 2: Persistent Approval via settings.json​

The Gemini CLI supports a persistent configuration file where you can explicitly list tools that should be auto-approved, making your choice permanent across all sessions and projects on your machine.

  1. Locate your User Settings File:

    • This file is typically located at: ~/.gemini/settings.json (on Linux/macOS)

  2. Edit the File: Open this settings.json file in a text editor.

  3. Add autoApprovedTools: You need to add a tools block with an autoApprovedTools array. For a command like npm, you are granting permanent permission to the underlying shell tool.

    If the file is empty or just contains a theme setting, it might look like this after your modification:

    {
      "theme": "Default",
      "tools": {
        "autoApprovedTools": [
          "shell"
        ]
      }
    }

    The "shell" tool is the one Gemini uses to execute commands like npm, git, ls, etc. By adding "shell" to this list, you are permanently instructing the CLI to auto-approve all shell commands without a prompt.

Any changes you make to the ~/.gemini/settings.json file will persist across all your Gemini CLI sessions until you manually remove the entry.

Adverts