Gemini CLI Documentation
by Google
The official manual for Google's open-source terminal coding agent — extensions, subagents, hooks, sandboxing and MCP.
Overview
Gemini CLI is Google's open-source terminal agent (Apache-2.0, github.com/google-gemini/gemini-cli, ~107k stars and ~14.5k forks, pushed daily as of 2026-09-09) and this is its official documentation site, organised into eight top-level sections. Get Started covers overview, quickstart, installation, authentication, a CLI cheatsheet and Gemini 3 support. Use Gemini CLI walks through file management, Agent Skills, context and memory, shell commands, sessions and history, task planning, Plan Mode, web tools, MCP setup and task automation. Features is the deepest section and documents extensions, Auto Memory, checkpointing, headless mode, git worktrees, hooks, IDE integration, MCP servers, model routing/selection/steering, notifications, subagents and remote subagents, rewind, sandboxing, settings, telemetry and token caching. Configuration covers custom commands, enterprise config, .geminiignore, model configuration, GEMINI.md, system prompts, themes and trusted folders; Reference covers the command, configuration and tools references plus keyboard shortcuts, the memory import processor and the policy engine. Separate Development and Releases sections document local development, integration testing, NPM package structure and release notes for both stable and preview channels. Extensions are the packaging unit: prompts, MCP servers, custom commands, themes, hooks, subagents and Agent Skills bundled behind a gemini-extension.json and installable from a GitHub URL.
At a Glance
- Topic
- Agentic
- Level
- All Levels
- Format
- Documentation
- Cost
- Free
- Duration
- ~2-3 hours to work through Get Started and Features; ongoing as reference
- Provider
- Hands-on
- Yes — code/exercises
- Certificate
- None
What You’ll Learn
- ✓Install and authenticate the CLI against a Gemini API key, Google account or enterprise config
- ✓Configure GEMINI.md context files, Auto Memory and the memory import processor
- ✓Register MCP servers and consume them as tools from inside the terminal agent
- ✓Package prompts, commands, skills, hooks and subagents as an installable extension
- ✓Run headless sessions and task automation so the agent works inside CI pipelines
- ✓Decompose long coding tasks with Plan Mode, subagents and remote subagents
- ✓Constrain execution with sandboxing, hooks, trusted folders and the policy engine
- ✓Control spend and latency through model routing, model steering and token caching
- ✓Recover work using checkpointing, rewind, session history and git worktrees
Highlights
- •Official Google documentation, not a third-party wiki — it tracks the Apache-2.0 repo that ships daily
- •Documents the security surface most agent-CLI docs skip: sandboxing, trusted folders and a policy engine
- •Release notes are published for both stable and preview channels, so you can see what changed before upgrading
- •Covers the extension format end to end, including writing and publishing your own to the gallery
- •A Development section exposes local dev, integration testing and NPM structure for contributors, not just users
Who It’s For
Best For
- ✓Engineers standardising one terminal coding agent across a team
- ✓Platform teams wiring internal MCP servers into a developer workflow
- ✓Developers comparing Gemini CLI against Claude Code, Codex CLI or OpenCode
- ✓Anyone automating repository chores headlessly from CI
Prerequisites
- •Comfort working in a terminal, plus Node.js and npm installed
- •A Google account or a Gemini API key for authentication
- •Basic git, since checkpointing and worktree features assume a repository
FAQ
What is Gemini CLI Documentation?
The official documentation for Gemini CLI, Google's open-source AI agent that runs in the terminal. It is written for developers who want a scriptable coding agent they can point at a repository, extend with MCP servers and skills, and constrain with a policy engine. After working through it you can install, authenticate, configure and automate the CLI, and package your own extensions for a team to install.
Is Gemini CLI Documentation free?
Gemini CLI Documentation is free to access.
What level is Gemini CLI Documentation for?
Gemini CLI Documentation is aimed at a all levels audience. Recommended background: Comfort working in a terminal, plus Node.js and npm installed, A Google account or a Gemini API key for authentication, Basic git, since checkpointing and worktree features assume a repository.
How long does Gemini CLI Documentation take?
Expect roughly ~2-3 hours to work through Get Started and Features; ongoing as reference. Most learners work through it at their own pace.
What will I learn from Gemini CLI Documentation?
You'll learn: Install and authenticate the CLI against a Gemini API key, Google account or enterprise config; Configure GEMINI.md context files, Auto Memory and the memory import processor; Register MCP servers and consume them as tools from inside the terminal agent; Package prompts, commands, skills, hooks and subagents as an installable extension; Run headless sessions and task automation so the agent works inside CI pipelines; Decompose long coding tasks with Plan Mode, subagents and remote subagents; Constrain execution with sandboxing, hooks, trusted folders and the policy engine; Control spend and latency through model routing, model steering and token caching; Recover work using checkpointing, rewind, session history and git worktrees.
Topics
Sources
This page was written from 2 sources, 1 on domains other than geminicli.com.