Google brings Gemini to the command line with an open-source AI agent

Gemini CLI, an open-source AI agent, gives developers lightweight access to its flagship model’s capabilities directly from the command line

Google is bringing its generative artificial intelligence (AI) model Gemini to the command line, a text-based computer interface that many developers consider home for its efficiency, ubiquity, and ability to get work done faster.

On June 25, the company introduced Gemini CLI, an open-source AI agent that gives developers lightweight access to its flagship model’s capabilities directly from the command line, also known as the terminal, to help them with a wide range of creative and productivity tasks.

The launch is Google's latest move to court developers with its AI offerings amid intensifying rivalry with technology firms such as OpenAI, Microsoft, and Meta, all of which are investing heavily to attract developers. India, which houses one of the world's largest developer bases, is a key part of these efforts.

Making tech more human-centric

Alan Hutchinson, Senior Director of Engineering at Google, said that infusing the terminal with AI is a major step towards making technology more human-centric.

"Computers have always required humans to adapt to the way computers work. With Gemini CLI, we are building a new era where computers are finally adapting to us," Hutchinson said.

He said the tech giant sees the command line as a key frontier where developers, and soon anyone, can get easy and direct help from Gemini, as the demand for powerful, integrated AI assistance continues to grow.

"We believe that Gemini CLI will fundamentally change how developers interact with AI. Gemini CLI removes the language barrier of cryptic commands and programming languages," he said.

Google is positioning Gemini CLI as a smart, interactive partner for developers to work with the tools on their computer. The agent, which is still in preview, can help them write code, debug issues, and streamline their workflow with natural language, it said.

A screenshot of Gemini CLI, an open-source agent for the command line

Hutchinson mentioned that the agent can read the entire project under the developer's supervision, understand dependencies, coding patterns, and function definitions to generate code that's tailored to their specific project.

Gemini Code Assist, the company's AI coding assistant, will also be built on the same foundational technology, giving developers access to these tools in code editors like Microsoft's VS Code.

"In VS Code, you can place any prompt into the chat window using agent mode, and Code Assist will work on your behalf to write tests, fix errors, build out features or even migrate your code. Based on your prompt, Code Assist’s agent will build a multi-step plan, auto-recover from failed implementation paths and recommend solutions you may not have even imagined," said Ryan J Salva, Senior Director of Product Management at Google.

Gemini CLI also comes with built-in support for the Model Context Protocol (MCP), system prompts, and configurable settings for both personal and team use to extend the agent’s capabilities.

Also Read: Google I/O 2025 was a litmus test. Did Sundar Pichai deliver?

Beyond coding

Gemini CLI also goes beyond coding, helping people with a wide range of tasks, from content generation and problem solving to deep research and managing tasks.

The company showed an example of a 30-second video of a cat visiting Australia, created using its image generation model, Imagen, and its text-to-video AI model, Veo.

Gemini CLI will be available free-of-charge for individual developers using Gemini Code Assist. One can log in with their personal Google account to get a free license that will give them access to an agent running Gemini 2.5, equipped with a massive 1 million token context window. People will be able to make up to 60 model requests per minute and 1,000 requests per day.

Professional developers who intend to run multiple agents simultaneously or use specific models can do so using Google AI Studio or Vertex AI key on a pay-as-you-go basis or get a Gemini Code Assist Standard or Enterprise license.

Gemini CLI’s code will also be publicly available on GitHub, allowing developers to inspect it, understand how it works, and assess its security implications, Salva said.