Google Gemini CLI: Commands, Setup, and Guide

Updated 5 November 2026 · Standard formula, shown below

Google Gemini CLI is an open-source command-line tool for asking Gemini to work with files and code in a terminal, with optional access to project context and tools. Install it with npm, authenticate, then run the gemini command; this guide covers setup, everyday commands, configuration, safety, and common fixes.

What Gemini CLI Is and How It Works

Gemini CLI brings a conversational AI workflow into a terminal. You can ask questions, request code changes, summarize files, or have it help investigate a project. It is designed to use the current directory as relevant context, rather than requiring you to paste every file into a web chat. Exact features and model availability can change, so check the project’s current documentation and release notes for details.

The CLI is not the same thing as a shell command that safely executes every suggestion. It can interact with local files and, when configured or approved, tools such as shell commands. Review proposed actions and changes before accepting them. Treat its answers as generated suggestions that may be incomplete or incorrect.

Install and Start a Session

The usual installation route is through npm, which requires Node.js and npm to be installed. Use the package name and instructions in the official Gemini CLI repository or documentation, since installation commands and requirements can change. A typical workflow looks like this:

Install a supported Node.js release and confirm that node and npm work in your terminal.
Install Gemini CLI using the current command documented by the project. A global npm install is commonly offered; alternatively, use a supported one-off package runner if the documentation provides one.
Change to the folder you want to work in, then run gemini.
Complete the authentication flow shown by the CLI. Depending on current options, sign-in may use a Google account or a configured API key or cloud project.
Read the displayed permissions and confirm only actions you intend to allow.

Starting the tool inside a project directory helps it understand which files matter. Avoid launching it from a broad directory such as your home folder if you do not want unrelated files considered as context.

Everyday Command Cheat Sheet

Gemini CLI has an interactive session as well as command-line options. Exact flags can differ by release; use gemini --help and the project documentation as the authority for your installed version.

gemini — start an interactive session in the current directory.
gemini --help — show available commands and options.
gemini --version — display the installed version when supported.
gemini "Explain this error" — pass an initial prompt from the command line, if accepted by the installed release.
/help — inspect interactive commands available in the current session.
/clear — clear session conversation context when available.
/quit or Ctrl+C — leave the session; key behavior may depend on the terminal and prompt state.

Slash commands are entered inside the interactive prompt, while shell commands are entered in your operating system’s terminal. Do not assume a slash command is a system command, or that every command listed in an older guide still exists.

Useful Prompts and Project Context

Specific requests produce more useful results than vague ones. State the goal, relevant files, constraints, and whether you want an explanation or edits. For example: “Inspect src/parser.ts, explain why empty input fails, and suggest a minimal fix. Do not edit files yet.” Once you have reviewed the suggestion, you can ask for a patch and inspect the resulting diff.

For larger tasks, divide work into steps: ask for a summary of the project, identify the likely files, request a proposed plan, then make and verify changes. Give the CLI useful context without exposing secrets. Do not paste passwords, access tokens, private keys, or confidential customer data into prompts.

Configuration, Tools, and Costs

Gemini CLI may support configuration files, extensions, model selection, and integrations with external tools. The available settings depend on the release and authentication method. Consult the current documentation before copying configuration examples: filenames, supported options, and defaults may change. Keep project-specific settings in the intended project configuration and personal settings in the user configuration, where the tool supports that distinction.

Some capabilities may invoke tools that read or modify files or run commands. Check which tools are enabled, what permissions they request, and whether each action needs confirmation. Extensions and third-party integrations can add capabilities but also introduce additional code and data-handling considerations; install only sources you trust.

There is no single cost that applies to every user. Usage may be governed by the sign-in method, account eligibility, service limits, or API and cloud billing terms. Check the current terms for the option you choose, including quotas and any charges, rather than assuming that installing the CLI makes all model usage free.

Common Problems and Safe Use

gemini is not found: confirm installation completed, then check that npm’s global executable directory is on your PATH. Restart the terminal after changing PATH.
Authentication fails: follow the sign-in prompt again, check network access, and verify that your selected account or API configuration is supported. Do not share credentials in a prompt or issue report.
It cannot find expected files: check the current working directory and ask it to inspect the specific relative path. Access can also be restricted by permissions or ignore rules.
Answers are inaccurate: provide the relevant error, file, and constraints; ask for evidence or a testable explanation, then verify the result yourself.
A change breaks the project: inspect the diff, use version control, and run the project’s tests or checks. Do not accept broad edits without reviewing them.

AI-generated code can contain bugs, unsafe assumptions, or outdated APIs. Keep backups or use version control before significant changes, avoid running destructive commands without understanding them, and review shell commands before approval. Never treat a generated response as proof that a program is secure or correct.

Frequently Asked Questions

Is Gemini CLI free?
Installing the command-line tool does not establish that all model use is free. Availability, quotas, and possible charges depend on the authentication and service option; check its current terms.

Do I need Node.js?
The npm installation route requires Node.js and npm. Follow the project’s current installation instructions for supported requirements and any alternatives.

Can Gemini CLI edit files?
It can assist with project changes when the relevant capability is available and permitted. Review the proposed edits and resulting diff before relying on them.

How do I see all available commands?
Run gemini --help in the terminal and use /help inside an interactive session. The installed version’s output is more reliable than an old cheat sheet.

Is Gemini CLI the same as Gemini in a browser?
No. The CLI is a terminal-based interface that can work with local project context and tools; a browser chat has a different interface and does not automatically share your local files.

Can I use it in any project folder?
You can start it from a folder you have permission to use, but choose the working directory deliberately. A narrower project folder helps avoid exposing unrelated local context.

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.