Google has introduced Gemini CLI, an open-source AI agent designed to run directly in the terminal. Aimed at developers who rely on the command line interface for its speed and portability, Gemini CLI brings the capabilities of the Gemini 2.5 Pro model to a familiar environment.
Currently available in preview, the new utility supports a wide range of tasks, from writing and debugging code to conducting research, managing tasks and automating workflows. It is also integrated with Google’s AI coding assistant, Gemini Code Assist, offering developers AI-first support in both VS Code and the terminal.
Designed for developers, with unmatched usage limits
Gemini CLI is free to use with a personal Google account, which grants access to Gemini Code Assist and the full capabilities of Gemini 2.5 Pro. This includes a 1 million token context window and what Google describes as the largest free usage allowance in the industry: up to 60 model requests per minute and 1,000 per day. These limits aim to support developers without interruption during the preview period.
Professional users who require more flexibility — such as simultaneous agent runs or model selection — can opt for usage-based billing via Google AI Studio or Vertex AI. Alternatively, they can subscribe to a Standard or Enterprise Code Assist plan.
Google has also highlighted the open-source nature of Gemini CLI, licensed under Apache 2.0. The company encourages the developer community to participate in its development by reporting bugs, suggesting new features and contributing code through its GitHub repository.
Features and integrations for real-world use
Gemini CLI is not limited to coding. Its capabilities include content generation, search-based prompt grounding, and file and command manipulation. Developers can automate tasks by running Gemini CLI non-interactively within scripts, or customise prompt behaviour using built-in configuration files.
The CLI is extensible via support for the Model Context Protocol (MCP) and comes with built-in extensions. Prompts can also be tailored through GEMINI.md and team settings, allowing users to create workflows suited to their own preferences.
According to Google, Gemini CLI can even be used to create content such as short videos. In one example, a prompt involving the adventures of a ginger cat in Australia can be turned into a short video using Google’s Veo and Imagen tools, showcasing the CLI’s broader utility beyond development work.
Powered by Gemini Code Assist
Gemini CLI shares its underlying AI technology with Gemini Code Assist, Google’s AI development assistant for coding in IDEs such as VS Code. Users can input prompts into a chat window in agent mode, allowing Code Assist to generate plans, recover from errors and suggest improvements automatically.
“Gemini Code Assist’s chat agent is a multi-step, collaborative, reasoning agent that expands the capabilities of simple-command response interactions,” Google said in the announcement.
Agent mode is available at no extra cost across all Code Assist plans — including free, Standard and Enterprise tiers — via the Insiders channel. According to Google, the free plan remains the most generous in terms of usage limits and takes under a minute to activate.
The Gemini CLI marks a further step in Google’s aim to embed AI into developers’ existing environments, making powerful tools available through familiar workflows and encouraging open-source contribution in the process.
