Gemini CLI is Google’s open-source AI agent for the terminal. Its README describes it as bringing the power of Gemini directly into your command line, and it is released under the Apache 2.0 license. You can run it without installing anything by using npx, or install it through npm, Homebrew, MacPorts, or Anaconda.
It works with Gemini 3 models and a 1M token context window, and it comes with built-in tools for Google Search grounding, file operations, shell commands, and web fetching. It supports MCP for custom integrations, and GEMINI.md context files let you tailor its behavior to a project. Because everything happens in the terminal, it suits developers who already work from the command line rather than inside an editor.
The free tier is the headline. Signing in with a Google account gives you 60 requests per minute and 1,000 requests per day with access to the Gemini 3 models. A Gemini API key also allows 1,000 requests a day, and Vertex AI is the enterprise route, with higher limits through a billing account. That is plenty for learning and side projects, but heavier daily use will push you towards an API key or a billed plan.
Verdict: the lowest-cost way to try an agent-first terminal tool from a major AI lab, and it is fully open source. If you would rather use Anthropic’s or OpenAI’s models, look at Claude Code or OpenAI Codex, and for the whole landscape see the best AI coding tools guide.
Source: the Gemini CLI repository, checked on September 20, 2026.
