Gemini CLI

Replicas supports Gemini CLI through its Agent Client Protocol mode.

Setup

Create a key in Google AI Studio, then add it under Gemini API Key in Coding Agents settings. See Credentials for personal and organization credential behavior.

Use Gemini CLI

Choose Gemini CLI from the workspace agent picker. Replicas runs gemini --acp with Gemini 3 Flash Preview by default. You can also select Gemini 3.1 Pro Preview. Gemini CLI streams messages and tool activity through ACP and supports persistent sessions, Plan Mode, image input, interruption, and queued-message steering. Replicas security policies run before each tool, and pull requests are discovered immediately after relevant tools complete. Inference analytics include Gemini CLI input and output token usage. Use Gemini CLI for repository tasks that benefit from Gemini’s coding and tool-use capabilities. Select another agent when the task depends on provider-specific features Gemini CLI does not support.

Programmatic use

Create a Gemini workspace through the API with "coding_agent": "gemini", or use the same value with the MCP create_replica tool.

{
  "name": "gemini-review",
  "environment_id": "ENVIRONMENT_UUID",
  "message": "Review this repository and fix the failing build",
  "coding_agent": "gemini"
}

For scheduled or event-driven work, choose Gemini CLI in the automation editor or create one with the CLI:

replicas automation create "gemini-review" \
  --environment ENVIRONMENT_UUID \
  --prompt "Review new pull requests for regressions" \
  --agent-provider gemini

See Automations, MCP, and the API guide for trigger and authentication details.

Get started

Workspaces

Coding agents

Integrations

Build and automate

Apps

Admin

Updates

Setup

Use Gemini CLI

Programmatic use

​Setup

​Use Gemini CLI

​Programmatic use

Setup

Use Gemini CLI

Programmatic use