Google finally did it.

They brought Gemini to the command line —

Which means you can now hand full control of your terminal over to AI.

It can write code.

Create files.

Execute commands.

And automate pretty much everything else.

Now, that used to be the kind of work programmers got paid for.

But once again, Google shows up late to the party.

Because let’s be honest — we already had Claude Code, OpenAI Codex, Warp, and dozens of other tools that have been dominating this space.

So what makes Gemini CLI interesting?

It has two big selling points:

1. It’s the only fully open-source mainstream solution.
2. It’s completely free, with a super generous usage limit — up to 1,000 model requests per day.

Now compare that to other top-tier coding models —

Some of them will run you $200/month, just to spit out piles of “meh” code in your terminal.

So yeah — you get what you pay for.

But is Gemini actually any good?

Let’s find out.

Today’s date is July 6, 2025.

You’re watching The Code Report.

As of today, Claude Code still leads the vibe-based coding scene.

They were the first major model provider to build a CLI tool like this.

Then OpenAI copied it.

Then Google copied them.

And now? You’ve got tons of third-party wrappers for every model you can think of.

But Claude has two major issues:

• It’s closed source
• It’s insanely expensive

You’ve probably seen those weekend hacker stories where someone spends thousands of dollars using Claude to build a side project — only to realize…

99% of those apps never make back the money spent on tokens.

Sad but true.

So what if you’re broke and can’t afford Claude?

That’s where Gemini comes in.

With 1,000 free model calls per day, it’s actually a decent alternative.

---

Getting Started with Gemini CLI

To begin, open up the official GitHub repo:

Make sure your environment is ready before installing:

• Node.js 18+
• A terminal environment (Command Prompt, Terminal, PowerShell, etc.)

To check your Node version:

node -v

On my machine, I’m running v20.0.0, so we’re good.

If yours is below 18, go to nodejs.org and update it.

Now for installation.

Google provides two install methods.

First, I tried:

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

But it didn’t work.

So I switched to:

npm install -g @google/gemini-cli

That worked.

Once installed, just type gemini in your terminal —

And follow the onboarding steps:

1. Theme selection — pick light or dark mode.
2. Authentication — You can log in with your Google Account (recommended, no setup needed),or use a Gemini API key if you need higher limits.

Once authenticated, you’re dropped into the Gemini CLI interface —

It’s a REPL-style environment, just waiting for your commands.

---

Hands-On: First Tasks

Let’s warm up with some basic tasks.

1. Check NYC Weather

Since Gemini CLI is a Google product, search is built-in.

I asked about the weather in New York — it went online, fetched the info, and showed me the result.

2. GitHub Code Summary

Next, I asked it to summarize code changes in a GitHub repo.

I used the Gemini CLI project itself.

First, I asked it to summarize yesterday’s commits — it said no changes were made.

Then I asked for updates from the past week — and it gave me a clean summary of the commits.

Not bad.

But this is just the beginning.

Before Gemini CLI even launched, we already had tools that could build full apps and websites using AI.

So I had to push it further.

---

Gemini CLI vs Cursor — Showdown

Let’s compare it head-to-head with Cursor — one of the most hyped coding tools globally.

Cursor uses Claude under the hood. I use it all the time to build websites.

Let’s see what happens when we give both tools the same prompt.

Task 1: Build a Tic Tac Toe Game

Step 1: Create a test/ directory.

Inside, make two folders:

• cursor_code/ — for code generated by Cursor
• gemini_code/ — for Gemini CLI output

Now open up both tools.

Here’s the prompt:

create a tik tac toe game with html5, javascript and css

Send the exact same prompt to both tools.

They’ll start executing commands on your machine —

Maybe running npm installs, maybe creating folders, files, etc.

Whenever I work with Vibe-style coding, I always let the tool automatically approve changes,

and give it permission to run commands freely.

Pretty quickly, Gemini CLI finished building the game.

Cursor took a little longer, but also completed soon after.

Let’s try both versions.

Functionally — they both worked.

But in terms of UI design, I have to say:

Gemini’s version felt a bit… rough.

Cursor’s output was just more polished.

---

Task 2: Build an Online Image Editor

New prompt:

create a modern web app that acts as an image editor
with contrast, brightness and hue controls.
tech stack: react

Send it to both tools.

Both tools started using React, and began installing dependencies with npm.

Same setup:

I let them run all commands, approve all changes.

And I noticed something cool —

Gemini CLI auto-corrected errors and fixed code on its own.

After a few minutes — Gemini CLI finished first again.

I tested its image editor — contrast, brightness, hue — all worked.

It even let me download the edited images.

Cursor finished two minutes later.

Same smooth experience.

I tested its version — it also handled all editing features perfectly.

So what’s the verdict?

Functionality-wise — both tools delivered.

But again — Cursor’s UI and UX were clearly a step ahead.

Final Thoughts

But let’s be real — Gemini CLI is already impressive.

It’s brand new. It’s open source. And it’s free.

From a cost perspective, this could save developers a ton of money.

If Google follows their usual playbook —

Gemini CLI will keep getting better.

It’ll gain adoption fast.

And who knows — in a few years, it might dominate and push other AI tools out of the market.

We’ll see.