Gemini CLI

Use Google's Gemini CLI with Verun — prerequisites, selecting the agent, and Verun-specific configuration.

Google Gemini CLI is Google's command-line coding agent powered by the Gemini model family.

Overview

Gemini CLI runs as a terminal process that Verun manages inside each task's worktree. It can analyze code, generate implementations, and run commands within the isolated worktree.

Prerequisites

Gemini CLI installed
A Google Cloud account with Gemini API access
Node.js 18+

Verify the CLI is available:

gemini --version

Selecting Google Gemini in Verun

Open a task from the task list.
Click Start Session.
In the session settings, select Google Gemini from the provider dropdown.
Choose a trust level.
Click Start.

Alternatively, set Gemini as the default in .verun.json:

{
  "agent": "google-gemini"
}

Configuration

Gemini CLI uses its own configuration for project-specific instructions. Place config files in your project root — they are available inside each worktree.

Verun-specific settings via .verun.json:

{
  "agent": "google-gemini",
  "setup": ["npm install"],
  "ports": [3000]
}

Known Limitations

Gemini CLI requires an active internet connection to communicate with Google's API.
The CLI must be installed and authenticated before starting a Verun session — Verun does not manage CLI authentication.