Gemini CLI is an Apache-2.0-licensed command-line interface that lets developers invoke the Gemini 2.5 Pro model (1 M-token context window) straight from their terminal.
It employs a reason-and-act (ReAct) loop backed by built-in tools (grep, file I/O, shell execution, Google Search, etc.) and supports external Model Context Protocol (MCP) servers for capabilities such as image/video generation with Imagen, Veo or Lyria.
Key features include:
- Code understanding & refactoring across large code-bases, bug-fixing, test generation and feature implementation.
- Workflow automation—invoke Gemini non-interactively inside scripts or CI pipelines to handle repetitive tasks.
- Extensibility & customization through user-defined tools, system prompts (
GEMINI.md) and per-team settings. - Shared technology with Gemini Code Assist—the same agent can run in VS Code or JetBrains IDEs, giving a consistent AI-first developer experience.
The preview release offers generous free quotas (60 requests/minute, 1 000/day) for individual Google accounts, while enterprise users can connect Vertex AI or AI Studio keys for higher-tier access.
Because the project is fully open source, the community is encouraged to inspect, extend and contribute to its evolving roadmap.
