Gemini CLI brings Gemini into the terminal as a coding and automation agent. In a project directory, you describe a task in plain language, such as explaining a codebase, fixing a failing test, or building a feature, and the agent reads files, proposes changes, runs shell commands, and iterates. It can search the web, use a context file to learn project conventions, and work with long contexts thanks to Gemini’s large context window.
It’s built to be extended. Support for the Model Context Protocol connects it to external tools, extensions package custom commands and integrations, and settings let teams control permissions and sandboxing. It’s open source under the Apache 2.0 license, and it can be scripted for use in automation and continuous integration. It also runs inside Gemini Code Assist’s agent mode in editors.
The tool is free with a personal Google account within usage limits, and paid plans or API keys raise limits. Like all coding agents, it can make mistakes, so review changes and be careful with commands that modify or delete files. Alternatives include Claude Code, OpenAI Codex CLI, Aider, Cline, and GitHub Copilot’s CLI. Gemini CLI suits developers who want a free, open-source terminal agent.