Google has released the AI agent Gemini CLI, a Google - version Cursor that can be used in all environments.

On June 25th, Google launched the open - source AI agent Gemini CLI (Command - Line Interface) today, integrating the capabilities of the Gemini large - model directly into the command - line interface that developers use daily.

This Gemini CLI, which can be very powerful in the terminal, combined with Gemini Code Assist deeply integrated in the IDE, can turn all the environments familiar to developers into Cursor. In any programming scenario, you can call Gemini to assist you in programming.

What can Gemini CLI do?

For developers, the command - line interface is not only a tool but also an important platform for daily work. Given its efficiency and universality, it has become the preferred environment for development work. As developers rely more and more on the command - line interface, the demand for integrating AI assistants is also increasing.

Gemini CLI provides a lightweight access method, allowing developers to interact directly with Google's AI models through the command line. It performs excellently in programming, but its functions are far from limited to this. This multi - functional local tool can be used in various scenarios such as content generation, problem - solving, in - depth research, and task management.

Google also combines Gemini CLI with its AI programming assistant Gemini Code Assist to ensure that whether they are free, standard, or enterprise users, they can enjoy the prompt - based AI programming experience in VS Code and Gemini CLI.

To use the free version of Gemini CLI, simply log in to your personal Google account to get a free license for Gemini Code Assist, and enjoy the Gemini 2.5 Pro version and a large context window of 1 million tokens. To ensure a smooth experience, Google also provides the highest free quota in the industry: a maximum of 60 requests per minute and a daily limit of 1000 requests.

For professional developers who need to run multiple agents simultaneously or users who prefer to specify models, they can implement pay - as - you - go through Google AI Studio or Vertex AI keys, or purchase the standard and enterprise versions of Gemini Code Assist.

Gemini CLI is currently in the preview stage and has powerful AI functions, including code understanding, file operations, command execution, and dynamic troubleshooting. It greatly enhances the command - line usage experience, enabling developers to write code, debug problems, and optimize workflows in natural language.

Its core advantage lies in the built - in multiple tools, including:

Use Google Search to provide real - time web content for prompts, giving the model rich external background information;

Support the Model Context Protocol (MCP) and a variety of extension plugins for easy function expansion;

Customize prompts and instructions to meet personalized needs and workflows;

Support non - interactive calls in scripts to achieve task automation and process integration.

Gemini CLI is completely open - source (under the Apache 2.0 license). Developers can view the code to deeply understand its principles and security mechanisms. Google encourages the global developer community to actively participate, submit vulnerability reports, feature suggestions, and code optimizations to jointly promote the development of the project.

In addition, Gemini CLI supports diverse customization based on MCP, system prompt files (GEMINI.md), and personal or team configurations, fully respecting the personalized needs of end - users.

Sometimes, the IDE (Integrated Development Environment) may be more suitable for specific tasks, but it also needs the support of a powerful AI assistant. Gemini Code Assist and Gemini CLI share the same technical underlying layer.

In VS Code, users can put any prompt into the chat window to start the agent mode. Code Assist will formulate a multi - step plan based on the prompt, automatically fix errors, write tests, expand functions, and even perform code migration. It can also recover from failed paths and recommend more innovative solutions.

The agent mode of Gemini Code Assist is open to all users and can be used for free through the Insiders channel. The free package has the highest usage quota in the market, and it is easy to start with almost no waiting.

An upgraded Copilot, a free Cursor

Behind the release of Google's Gemini CLI actually hides a philosophical debate about the future: Is it to create a brand - new AI tool to "replace" developers' old habits, or to "inject" AI capabilities into developers' existing workflows?

AI - native editors represented by Cursor have chosen the former. They boldly claim: "Come to me, and I've built a brand - new studio from top to bottom for AI." The advantage of this approach is that the integration of AI and the editor can be seamless, and some disruptive functions can be achieved. But the cost is that developers need to leave their familiar and highly customized comfort zones to adapt to a new environment.

Google has given a completely different answer with the Gemini series. It has chosen a path of "walking with you" and says to developers: "You don't need to change anything. Your favorite VS Code, the entire JetBrains suite, and the command line you can't live without... They are all still there. We just hand over the most powerful AI to you in the form of 'plugins' and 'commands' to transform your old friends."

Google's bet is clear: it believes that the "muscle memory" and emotional attachment of the vast majority of developers to their tools are strong. It doesn't intend to reshape your world but to make you omnipotent in your own world.

This competition in AI development tools is becoming more and more exciting.

This article is from "Tencent Technology", author: Jin Lu. Republished by 36Kr with permission.