Gemini CLI Docs

Gemini CLI brings the full power of Gemini models straight to your terminal. Whether you're prototyping in a REPL, building custom tools, or automating workflows, Gemini CLI helps you integrate AI directly into your development environment—with speed, clarity, and flexibility.

This guide walks you through installation, usage, architecture, and extension. It's written for developers who want more than just API references—clear, actionable documentation built for real-world application.

---

What Is Gemini CLI?

Gemini CLI is a local-first command-line tool that connects your terminal with Gemini’s AI models. It provides:

• An interactive REPL experience for real-time model exploration
• A flexible client-server architecture for tool integration and request management
• A growing ecosystem of built-in tools for file I/O, shell execution, web access, and memory storage

The CLI is divided into two packages:

• packages/cli – the user-facing command-line interface
• packages/core – the local server that handles tool execution and API communication

This design lets you build, test, and scale intelligent workflows right from your terminal—no external services required beyond Gemini.

---

Highlights at a Glance

• Interactive AI REPL – Quickly test ideas and prompts with real-time feedback
• Tooling System You Can Extend – Add custom commands or tweak existing tools
• Clean Modular Design – Separate concerns between CLI and core logic
• Works Entirely Locally – Ideal for development, scripting, and experimentation

---

Getting Started

Ready to dive in? The Execution and Deployment section includes:

• How to install dependencies and set up your environment
• Steps to run Gemini CLI locally
• Connecting your setup to the Gemini API

The process is smooth for developers on macOS, Linux, or Windows running Node.js.

---

How It Works: Under the Hood

Gemini CLI operates on a client-server model:

• The CLI (packages/cli) handles user input and interfaces with the REPL
• The Core (packages/core) interprets commands, routes them to tools or Gemini, and returns results

This separation allows you to:

• Test tools independently
• Run long or async tasks with less friction
• Connect other systems or UIs beyond the terminal

For visuals and architecture diagrams, explore Architecture Overview and Core Introduction.

---

Using Gemini CLI

Core Commands

From querying models to running scripts, Gemini CLI includes commands like:

• ask – Send questions or prompts directly to a Gemini model
• run – Execute a tool or code block
• config – Adjust local settings and preferences

See Commands for syntax, examples, and options.

Tools API

Gemini CLI includes a modular tools API with functions that simplify automation:

• File Tools: read_file, write_file, read_many_files
• Shell Tools: Execute shell commands directly within workflows
• Web Tools: Use web_fetch or search via google_web_search
• Memory Tools: Save and recall model memory for contextual tasks

Each tool is fully documented in the Tools section with practical examples.

Configuration

Customize your CLI experience with a local config file—control API tokens, output styles, and telemetry preferences. Learn more in Configuration.

---

Advanced Usage

Checkpointing

Need to pause and resume work? Use the checkpointing system to save model interactions and restore them later. Great for long prompts, iterative workflows, or session-specific prototyping.

Extend the CLI

You can build your own tools and integrate them into Gemini CLI. Visit the Extensions section to learn how to:

• Create a new tool
• Register it with packages/core
• Surface it via the CLI

Telemetry

Anonymous usage data helps us improve Gemini CLI. Telemetry is optional and can be disabled in your config. See Telemetry for full details.

---

Developer & Contributor Guide

Want to contribute? Head over to Contributing & Development Guide for:

• Overview of the codebase and project structure
• How to set up your local development environment
• CI practices, testing strategies, and code standards

New contributors are welcome—modular code makes onboarding fast and fun.

---

Troubleshooting

Ran into an issue? The Troubleshooting Guide offers:

• Solutions to common errors
• Debugging tips for tools and CLI behavior
• A list of known issues with current workarounds

For bugs or feature requests, open an issue on GitHub.

---

FAQs & Policies

See our Terms of Service and Privacy Notice for transparency around data usage, open source licensing, and community guidelines.

---

Who It’s For

Gemini CLI is designed for:

• Developers integrating AI into local tooling or workflows
• Data Scientists testing prompts and automation logic
• DevOps Engineers scripting repeatable tasks using shell and file tools
• Open Source Contributors looking to extend or adapt the system

---

Final Thoughts: Why Choose Gemini CLI?

Gemini CLI isn’t just an API wrapper. It’s a robust, developer-friendly interface that brings Gemini’s capabilities into your everyday terminal workflow.

• Modular architecture built for customization
• REPL-based interaction for real-time experimentation
• Extensible tooling to fit any dev stack

Whether you're building, testing, or just exploring—Gemini CLI gives you everything you need to work smarter with AI, right at your fingertips.