Gemini CLI
Google's open-source Gemini agent for the terminal
About Gemini CLI
Gemini CLI is Google's open-source AI agent for the terminal. Released under the Apache 2.0 license and developed in the open on GitHub, it brings Gemini models into your command line so you can query and edit large codebases, generate new apps from PDFs, images or sketches, debug problems in natural language and automate operational chores such as querying pull requests or handling complex rebases. It works with your local project context: it reads and writes files, runs shell commands and can search and fetch pages from the web, all from a single gemini command.
You install Gemini CLI with npm, Homebrew or MacPorts, or run it once with npx. It needs Node.js 20 or later and supports macOS 15+, Windows 11 24H2+ and Ubuntu 20.04+, and it comes pre-installed on Google Cloud Shell and Cloud Workstations. You then sign in with Google using your organization's Code Assist license and Google Cloud project, or supply a Gemini API key or Vertex AI credentials.
Inside a project you run gemini and describe the task.
- •Plan Mode - A read-only environment for researching the project, evaluating trade-offs and settling on a strategy before any code is modified, entered with Shift+Tab or the /plan command.
- •Subagents - Specialist agents with their own system prompt, tools and context window.
- •Checkpointing and Restore - An opt-in setting that, before AI edits, snapshots your project in a shadow Git repository in your home directory, and /restore reverts the files and the conversation.
- •Sandboxing and Trusted Folders - Shell commands and file changes can run inside a Docker or Podman container that limits access to the project directory, and trusted folders set execution policies per folder.
- •Extensions, Skills and MCP - Connect MCP servers, install extensions, add Agent Skills, hooks and custom slash commands, and keep persistent project instructions in GEMINI.md files.
Gemini CLI now suits engineering teams on Google Cloud that hold Gemini Code Assist Standard or Enterprise licenses, which come with 1,500 or 2,000 model requests per user per day, or that use Code Assist for GitHub. Platform and DevOps engineers can use its headless mode to automate repository and cloud chores, and developers with paid Gemini API or Vertex AI keys can keep using it on a pay-per-token basis.
Pricing
- Free$0/mo
- Gemini Code Assist Standard$22.80/mo
- Gemini Code Assist Enterprise$54/mo
