Skip to main content
The gemini runner invokes the Google Gemini CLI binary as a subprocess. This is the easiest way to get started without any API costs — Google’s free tier provides a generous number of requests per day with no credit card required.
The Gemini runner is currently in beta. Output parsing edge cases may exist. If you encounter unexpected behavior or formatting issues, please open an issue on the GitHub repository.

Requirements

  • The gemini CLI installed globally:
    npm install -g @google/gemini-cli
    
  • A Google account for authentication. The free tier does not require a credit card.
  • Node.js 20 or later.
Verify the binary:
gemini --version

Authentication

The Gemini CLI uses Google OAuth for authentication. On first run, it opens a browser window to complete the login flow:
gemini auth login
After authenticating, credentials are cached locally. The agent uses these cached credentials when invoking the Gemini CLI — no API key needs to be entered during remoteagent init.

Setup

remoteagent init --runner gemini
Or choose gemini when prompted by the interactive wizard. Since no API key is required for the free tier, the wizard will skip the API key prompt.

How it works

The runner executes:
gemini "<prompt>"
Output is captured from stdout and streamed back to the Redis channel. The Gemini CLI writes its response progressively, so Telegram receives chunks as they arrive.

Free tier limits

Google’s free tier for Gemini CLI provides a daily quota of requests. For most development tasks — reviewing a file, fixing a bug, writing tests — you are unlikely to hit the limit during normal usage. If you need higher throughput or want to use Gemini via an API key (for the Gemini API), you can set the GEMINI_API_KEY environment variable before starting the agent. The gemini CLI will use the API key instead of the OAuth flow.

Pros and cons

ProsCons
Free tier — no credit card requiredBeta status — some output parsing edge cases
No API key to manage for free usageDaily rate limits on the free tier
Access to Gemini 2.x modelsSubprocess overhead per invocation
Less capable than Claude for complex coding tasks (varies by task)

Updating the binary

npm update -g @google/gemini-cli
gemini --version
Restart the agent after updating:
remoteagent start