Blogs/AI

What is Google Gemini CLI & how to install and use it?

Written by Sharmila Ananthasayanam
Jul 29, 2025
2 Min Read
What is Google Gemini CLI & how to install and use it? Hero

Ever wish your terminal could help you debug, write code, or even run DevOps tasks, without switching tabs? Google’s new Gemini CLI might just do that.

Launched in June 2025, Gemini CLI is an open-source command-line AI tool designed to act like your AI teammate, helping you write, debug, and understand code right from the command line.

What is Gemini CLI?

Gemini CLI is a smart AI assistant you can use directly in your terminal. It’s not just for chatting, it’s purpose-built for developers.

Whether you're reviewing code, fixing bugs, managing Git workflows, or generating docs, Gemini CLI helps you get it done faster. It uses a clever ReAct loop (Reason + Act) to make decisions. This means it doesn’t just give answers, it reasons, acts, and interacts with both your local system and remote model context servers (called MCPs) to perform tasks intelligently.

How to Get Started with Gemini CLI?

Before you begin, make sure you have Node.js v18 or higher installed on your system.

Option 1: Use without installing

You can try it instantly using:

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

No setup required.

Getting Started with Google Gemini CLI
Learn how to install and use Google Gemini CLI for model interaction, scripting, and automation.
Murtuza Kutub
Murtuza Kutub
Co-Founder, F22 Labs

Walk away with actionable insights on AI adoption.

Limited seats available!

Calendar
Saturday, 11 Apr 2026
10PM IST (60 mins)

Option 2: Install it globally

If you plan to use it often, install it globally:

sudo npm install -g @google/gemini-cli

After installation, just type gemini in your terminal to launch it. 

Installing Gemini globally

Select your desired theme and click Enter.

Authentication Options for Using Gemini CLI

Once Gemini CLI starts, it’ll ask you to authenticate. You’ve got three options:

  1. Login with Google
    • 60 requests per minute
    • 1000 requests per day
  2. Use Gemini API Key
    • Either create a.env file in your project folder and add the line: GEMNI_API_KEY=<your_gemini_api_key> to use your Gemini API key or just enter it in your terminal directly: 

> export GEMINI_API_KEY=<your_gemini_api_key>

  • Free Tier:
    • Flash model only
    • 10 requests/min
    • 250 requests/day
  • Paid Tier:
    • Quota depends on your tier
  1. Use Vertex AI
    • Available for paid accounts
    • Quotas depend on your tier

What Can Gemini CLI Do?

Gemini CLI is more than just an AI chatbot. Here are some of its coolest features:

  • Understand Big CodebasesAsk it questions about your entire codebase, even beyond a 1 million token context window.
  • Generate Apps from PDFs or Sketches Leverage Gemini’s multimodal power to create code from designs or documents.
  • Automate DevOps Tasks Handle pull requests, resolve complex rebases, or run scripts, all with a single prompt.
  • Extend with Tools and MCP Servers Connect tools like Imagen, Veo, or Lyria to add media generation.
  • Ground Queries with Google Search Tap into the world’s knowledge using the built-in Google Search tool.
Getting Started with Google Gemini CLI
Learn how to install and use Google Gemini CLI for model interaction, scripting, and automation.
Murtuza Kutub
Murtuza Kutub
Co-Founder, F22 Labs

Walk away with actionable insights on AI adoption.

Limited seats available!

Calendar
Saturday, 11 Apr 2026
10PM IST (60 mins)

Helpful Tips

  • Press / anytime to open the command helper.
  • Type /docs and hit Enter to open the official Gemini CLI documentation in your browser.

Final Thoughts

Gemini CLI is a smart and flexible tool that helps with coding tasks. It works great for small personal projects or large teamwork, making it easier to write, fix, and manage code right from your terminal.

What makes it stand out?

  • Open-source transparency
  • A generous free tier
  • Advanced AI from Google Gemini
  • Deep integration with developer tools
  • Multimodal support (text, code, images, and more)

And most importantly, it runs in the terminal, your favourite place to work.

Author-Sharmila Ananthasayanam
Sharmila Ananthasayanam

I'm an AIML Engineer passionate about creating AI-driven solutions for complex problems. I focus on deep learning, model optimization, and Agentic Systems to build real-world applications.

Share this article

Phone

Next for you

Cost to Build a ChatGPT-Like App ($50K–$500K+) Cover

AI

Apr 7, 202610 min read

Cost to Build a ChatGPT-Like App ($50K–$500K+)

Building a chatbot app like ChatGPT is no longer experimental; it’s becoming a core part of how products deliver support, automate workflows, and improve user experience. The mobile app development cost to develop a ChatGPT-like app typically ranges from $50,000 to $500,000+, depending on the model used, infrastructure, real-time performance, and how the system handles scale. Most guides focus on features, but that’s not what actually drives cost here. The real complexity comes from running la

How to Build an AI MVP for Your Product Cover

AI

Apr 7, 202613 min read

How to Build an AI MVP for Your Product

I’ve noticed something while building AI products: speed is no longer the problem, clarity is. Most MVPs fail not because they’re slow, but because they solve the wrong problem. In fact, around 42% of startups fail due to a lack of market need. Building an AI MVP is not just about testing features; it’s about validating whether AI actually adds value. Can it automate something meaningful? Can it improve decisions or user experience in a way a simple system can’t? That’s where most teams get it

AutoResearch AI Explained: Autonomous ML on a Single GPU Cover

AI

Apr 2, 20268 min read

AutoResearch AI Explained: Autonomous ML on a Single GPU

Machine learning experimentation sounds exciting, but honestly, most of my time goes into trial and error, tuning parameters, rerunning models, and figuring out what actually works. I’ve seen how slow this gets. Some reports suggest up to 80% of ML time is spent on experimentation and tuning, not building real outcomes. That’s exactly why AutoResearch AI stood out to me. Instead of manually running experiments, I can define the goal, give it data, and let an AI agent continuously test, evalua