Google King Explodes Open Source Programming Agent: Gemini 2.5 Pro Stuffs Command Line, 1000 Free Calls a Day, Redefines CLI Development Experience

1. Introductory section

Instead of buying Claude Code and OpenAI Codex CLI, Google's big benefactors are here to give developers another bonus!

Just now, Google officially open sourced an AI programming agent called Gemini CLI.

This is not another simple chatbot, but a "universal terminal Copilot" that can deeply integrate into your workflow, understand your code, connect to your tools, and be driven by you almost indefinitely.

2. Features of Gemini CLI

Google's move can be described as full of sincerity, and even some "radical":

• Completely free and open source: based on Apache 2.0 protocol, the code is publicly available, and developers around the world are welcome to build together.

• The most generous free quota in history: Individual users who log in through Google account can get the right to use Gemini 2.5 Pro, enjoy the context window of 1 million Token, and the quota of 60 requests per minute and 1000 requests per day. This can be called the "ceiling" level in the industry, and can be called a great benefactor.

• It's more than just programming: it's a generalist, from code migration, Debug, to content generation, deep research, and even the ability to connect to MCP to call Veo and Imagen models in one sentence.

• Native integration and seamless connection: It not only exists in your terminal, but also shares the underlying technology with Gemini Code Assist in VS Code, allowing AI capabilities to flow seamlessly between your IDE and CLI.

"For developers, the command-line interface (CLI) is not just a tool, it's home," said Ryan J. Salva, Google's senior director of product management, in a blog post. We launched Gemini CLI to bring the power of Gemini directly into the home.

Who is this Gemini CLI with high expectations? How will it change our daily routine of "typing code, fixing bugs, and doing operation and maintenance" in the dark terminal?

For the first time, we have brought you the most detailed and in-depth interpretation of the whole network.

3. Redefining the developer's home: why CLI?

In an age of GUIs, why would Google go "back to basics" and create a AI Agent? Tailored to the oldest developer tool of all: the command line?

The answer is simple: for real developers, the terminal is never out of date.

Its efficiency, ubiquity, and portability make it the go-to tool for getting the job done. From git version control, to SSH remote management, to docker container orchestration, countless core development tasks occur in this small black window. Developers' dependence on terminals is enduring, and the demand for integrated AI assistance is also rising.

Before that, if developers wanted to use ai in CLI, they often needed to use complex API calls, install third-party packaging libraries, or constantly switch between browsers and terminals, and the experience was very fragmented.

The birth of Gemini CLI is to completely break this barrier. Its core idea is to provide developers with the most direct path from Prompt to model, so that AI can become a "first-class citizen" of the terminal environment.

It is not an "alien species" that you need to learn and adapt to, but a "native assistant" that can understand your natural language, understand your current working directory, and directly operate your local files and tools.

You can use one sentence to make it:

• "Summarize all the code changes on the main branch yesterday."

• "Help me convert all.jpg images in this directory to PNG format and rename them with the EXIF date of the photo."

• "Read this PDF requirements document to help me generate a project skeleton with React and Tailwind CSS."

This seamless integration of big-model capabilities into the core workflow experience is what Google calls "redefining the developer's home.".

4. The Most Generous Ever: Gemini 2.5 Pro Is Free to Use

Of course, for developers, fancy features are not as real as "free" and "powerful.".

At this point, Gemini CLI offers the industry's top "free luxury package".

As long as you have a personal Google account, you can automatically get a free Gemini Code Assist license without binding a credit card. With this license, you can directly call one of Google's most powerful current models, Gemini 2.5 Pro, in the Gemini CLI.

This means that you will have:

1. Huge context window: Up to 1 million Tokens of context processing capabilities allow you to "feed" entire code bases and complex documents to it for in-depth analysis and understanding.

2. Amazing amount of free calls: 60 requests per minute, 1,000 requests per day. The official blog post proudly declares that this is to "ensure that you hardly, if ever, reach the upper limit during the preview.". This amount far exceeds the free package of most similar products on the market, which is enough to meet the daily high-intensity use of most individual developers and enthusiasts.

Gemini CLI also provides flexibility if you are a professional developer who needs to run multiple Agents at the same time, or if you want to use a specific model. You can use Google AI Studio or Vertex AI's API Key for Pay-As-You-Go, or upgrade to Gemini Code Assist's Standard or Enterprise licenses.

This pricing strategy of "free entry, high ceiling and professional options" will undoubtedly greatly reduce the threshold for developers to experience cutting-edge ai technology, and also demonstrate Google's determination to quickly build a large developer community.

5. More than just programming: a truly scalable "all-purpose Agent"

If Gemini CLI is just a chat tool that can call Gemini 2.5 Pro for free, it is not enough to be called "King Bomb".

Its real power lies in its inherent design as an "Agent"-it has tools that can be invoked and the ability to be extended.

The power of the Gemini CLI comes from the built-in capabilities of several cores:

1. Google Search real-time information access: The Google search tool is built in, which can get the latest information online at any time and provide real-time and external context for the model. This means that if you ask it about a new framework that has just been released, it will not be ignorant because of the training data deadline.

2. Local file system interaction: It can read and write your local files and directories like an experienced developer. You can easily inject the contents of a file or entire directory into your Prompt using the @ symbol, such as the @ SRC/components/Explain this React component. More intelligently, it follows the rules of.gitignore by default, automatically filtering out extraneous files such as node _ modules/, dist/, and focusing on your core code.

3. Shell command execution: Pass! Prefix, you can execute any system Shell command directly inside the Gemini CLI, or simply switch to "Shell mode", operate the terminal as usual, and seamlessly switch back to talking to the AI when needed. This provides a perfect blend of AI assistance and traditional CLI operations.

4. Extensible Tool Community (MCP): Gemini CLI supports the emerging Model Context Protocol (MCP). This is a central but important feature that gets a passing mention in the official blog post. Simply put, the MCP allows the Gemini CLI to connect to an external "tool server," thereby extending its capabilities indefinitely. You can build your own MCP server to encapsulate the company's internal APIs, database query tools, project management suites (such as Jira) into AI-callable tools. MCP will bring endless imagination space to Agent.

5. Customizable "Layered Memory" (GEMINI. MD): This is another very deep customization feature. You can create Markdown files called GEMINI. MD in your user home directory, project root directory, or even subdirectories. Instructions, background information, or code snippets written in these files are automatically loaded by Gemini CLI and become the "long-term memory" or "System Prompt" of the model. This hierarchical design means that you can set a global behavior directive for all projects, provide a specific context for a particular project, and even specify more granular rules for a module within a project. Through the/memory series of commands, you can view and refresh these memories at any time, allowing the AI to act exactly according to your specific needs and workflow.

It is these built-in and extensible tools that transform Gemini CLI from a "conversational model" into an "agent" capable of perceiving the environment, performing tasks, and solving problems.

6. Getting Started: Installing the Gemini CLI into Your Terminal

Having said all that, it's time to experience it with your own hands. Installing and starting the Gemini CLI is a simple process.

Step 1: Ensure the environment

You need to install a Node. JS (version 18 or later). You can check your version by typing node -v in the terminal.

Step 2: Run the installation command

Open your terminal and execute the following command:

npx https://github.com/google-gemini/gemini-cli

Alternatively, you can choose to install globally so that you can use the Gemini command directly anywhere:

npm install -g @google/gemini-cli

On first run, it will prompt you to choose a preferred theme color.

Step 3: Certification

Next, it prompts you to log in to your personal Google account through your browser.

This process is based on OAuth 2.0 and is very secure. After logging in successfully, your terminal will automatically get the aforementioned free and high-quota access to Gemini 2.5 Pro.

Step 4: Start the conversation!

Now you're ready to talk to Gemini.

You can enter a brand new project directory and let it help you start from scratch:

$ mkdir my-discord-bot && cd my-discord-bot

$ gemini

> Write me a Gemini Discord bot that answers questions using a FAQ.md file I will provide

Or, tap into an existing code base and let it help you understand and adapt the project:

$ git clone https://github.com/google-gemini/gemini-cli

$ cd gemini-cli

$ gemini

> Give me a summary of all of the changes that went in yesterday

> Help me migrate this codebase to the latest version of Java. Start with a plan.

It's that simple.

7. Master advanced: must know the CLI command book

To help you maximize the power of the Gemini CLI, we've compiled a list of commands from the official documentation. Master them and your terminal operating efficiency will take off again.

These commands fall into three main categories, which are distinguished by different prefixes:

1. Slash (/) command: Session and meta control

These commands are used to control the behavior of the CLI itself, managing sessions and settings.

• /help or/?: Displays help information and lists all available commands.

• /chat save: Save the current conversation history with a tag to facilitate subsequent recovery.

• /chat resume: Resumes a conversation from one of the previously saved tabs.

• /chat list: Lists all saved chat tags.

• /compress: Replace the entire chat context with a summary. After a long conversation, this command can help you save a lot of Tokens while preserving the core information.

• /memory show: Displays the hierarchical memory contents currently loaded from all GEMINI. MD files.

• /memory refresh: Reloads memory from the GEMINI. MD file, useful when you have modified the configuration file.

• /restore [tool _ call _ ID]: Undoes the file modifications made by the last tool execution. This is a "regret medicine" feature that needs to be turned on with the --checkpointing flag when using the Gemini command.

• /stats: Displays detailed statistics of the current session, including Token usage, session duration, etc.

• /theme: Opens the theme selection dialog.

• /clear (Ctrl + L): Clear the terminal screen.

• /quit or/exit: Exit Gemini CLI.

2. At (@) command: Inject file and directory context

This is the core functionality for discussing code and documentation with ai.

• @: Inject the content of the specified file into your Prompt. For example: What is this file about? @README.md

• @: Inject the content of all text files (not ignored by gitignore) in the specified directory and its subdirectories. For example: @src/my_project/ Summarize the code in this directory.

• Spaces in the path need to be \ escaped with a backslash.

3. Exclamation (!) Command: Seamlessly interact with your Shell

This allows you to execute system commands without having to exit the Gemini CLI.

• !<shell_command>：Execute a single Shell command and return to the Gemini CLI. For example: !ls -la or !git status.

• ! (Enter separately): Switch to Shell Mode. In this mode, anything you type is executed directly as a Shell command, and the terminal prompt changes color to distinguish it. Enter again! You can exit Shell mode and return to the conversation with the AI.

By using these commands skillfully, you can truly appreciate the power of Gemini CLI as a "terminal native AI".

8. Strategic linkage: the combination of CLI and IDE

Google's chess game is bigger than we thought.

Gemini CLI is not a stand-alone product. In its blog post, Google specifically emphasized that it shares the same underlying technology as Gemini Code Assist in VS Code.

This means that the powerful "Agent Mode" you experience in the VS Code chat window has exactly the same core driving force as the Gemini CLI. This Agent pattern gives you the ability to build multi-step plans, automatically recover from failed execution paths, and come up with solutions you may never have thought of.

The intent of this linkage strategy is clear: Google wants to provide developers with a unified, coherent ai-assisted experience, whether you're in a feature-rich IDE or a streamlined CLI.

• Gemini Code Assist in the IDE is your best partner when you need to do complex refactoring tasks that require frequent viewing and editing of code.

• When you need fast code queries, batch tasks, automated scripting, or remote server operations, the Gemini CLI in the CLI provides the most direct and lightweight help.

They are not competitive relationships, but two sides of the same body, which together constitute an AI tool matrix covering the main work scenarios of developers. More importantly, the high quota of the free plan also applies to Gemini Code Assist, and developers can switch freely between the two tools without worrying about the cost.

9. The Future of Open Source: A Community-Driven AI Agent

Finally, we must re-emphasize the open source nature of Gemini CLI.

Google's choice to open source such a core tool under the Apache 2.0 protocol is a strong signal in itself. This means:

• Transparency and security: Any developer can examine its source code, understand how it works, verify its security, and address concerns about data privacy.

• Community building: Google made it clear that it "fully expects (and welcomes) the global developer community to contribute to this project by reporting bugs, suggesting features, continuously improving security practices, and submitting code improvements.". An active community will be the source of its vitality.

• Unlimited customization and forking: If you are not satisfied with some of the features of Gemini CLI, or want to build a more professional tool for a specific field based on it, you can fork a copy of the code and make deep magic changes.

It is foreseeable that a thriving ecosystem will emerge around Gemini CLI. We may see:

• Community-contributed MCP tool servers that connect to various popular SaaS services.

• GEMINI. MD configuration templates optimized for a particular programming language or framework.

• More creative third-party endpoint applications based on Gemini CLI core competencies.

All in all, Google's Gemini CLI is more than just a new tool. It is more like a declaration, announcing a new paradigm of developer workflow in the era of AI Agent.

In an extremely "down-to-earth" way, it goes deep into the command line "home" that developers are most familiar with, supplemented by an unprecedented free strategy and a thorough open source spirit, and sends a sincere invitation to developers around the world:

Come on, let's rebuild our workspace with ai.

So, what are you waiting for? Upgrade your terminal intelligence development experience now.

10. The latest version of DeGPT has integrated Gemini 2.5 pro

The latest DeGPT has inherited Gemini 2.5 Pro and fully supports the above functions and operations.

At the same time, DeGPT also integrates dozens of mainstream models, which is truly the king of AGI.

With its complete functions and favorable price, it can be called the most AI in 2025.

Learn more: https:www.degpt.ai

GitHub repository address: https://github.com/google-gemini/gemini-cli