Google LLC today introduced a new way for developers to work with Gemini, the company’s flagship artificial intelligence model, at the command line using a customizable agentic tool.

The command line interface is the text-based way for users to interact with a computer’s operating system. To many users, it’s seen as an almost mystical method for connecting to the computer, but for developers, it’s the rawest method of installing, modifying and using commands.

However, it’s also the most fraught with risk. A single typo can cause a command to fail. Command line lacks the ease of icons and graphics; users must type commands directly into a terminal. It’s the mainstay of hacker movies where actors type furiously at their keyboards and text flies across the screens.

Gemini CLI is an open-source AI agent that brings the power of Gemini directly into the terminal. It acts as a lightweight interface for the AI model that allows users to prompt it to take actions using plain English language: No more typing out long, obtuse commands, looking up documentation in books or on the web and trying to recall long-forgotten scripts.

Now users can have Gemini execute file modifications, run long sequences of commands to generate or manipulate content, install or migrate code, do code analysis and more.

Gemini can perform all of these tasks and self-correct when needed.

“Computers have required users to adapt to how computers work, now computers will adapt to them,” said Taylor Mullen, senior staff software engineer at Google. “We think builders, developers and creators will love this. This will be very approachable. With this, the command line is no longer for rigid commands; this tool changes the command line from simple commands to an active partner.”

For example, if a user wants to run a series of commands that are interconnected but isn’t sure about all the necessary sub-parameters to activate them, many commands offer a “–help” parameter. This parameter provides information on how the command works. By prompting the Google CLI with the user’s desired actions, the model can examine the command-line tools (including the help text) to ensure they”re used correctly.

Gemini CLI is capable of providing code understanding, file manipulation and command execution and dynamic script troubleshooting at the command line. It’s grounded with Google Search to obtain extra information about commands, scripts and information on the computer and users can extend its capabilities with the Model Context Protocol.

Users can also provide system prompts for Gemini CLI in a file named “GEMINI.md” and settings for both personal and team configurations so the model can be tailored to project standards. A system prompt is a set of overarching instructions that always runs and acts as a foundational guide to shape how an AI responds. It can be used to set project conditions, best practices, point it at favorite codebases and personalization.

Developers can use Gemini CLI free of charge today by logging in with their personal Google account. This provides up to 60 model requests per minute and 1,000 requests per day for free alongside access to Google 2.5 Pro. Professional developers who want to run multiple agents simultaneously, or prefer to use specific AI models, can use Google AI Studio or Vertex AI keys for usage-based billing.

Images: Google