Stop Hitting Paid Rate Limits: My Free Gemini CLI Dev Setup

Just today, I received an email from Claude regarding the introduction of weekly usage limits. Following the change in price plan for Cursor (and drama), it's not surprising to see Claude adjust their rate limits as well. Anthropic initially offered generous limits to accelerate the adoption of Claude Code, but now it looks like they’re reeling in. I wouldn’t be shocked to see tighter limits and some community pushback in the weeks ahead.

As such, let's take a look at the alternatives, away from Cursor, Claude Code, and all those paid AI coding assistants. We’ll also integrate with an AI autonomous coding agent to further enhance its capability, pushing towards truly an alternative to Cursor and Claude Code!

Gemini to the rescue

Jun 25, 2025: Google dropped the bomb on their answer to Anthropic's terminal cli dominance: Gemini CLI. As it currently stands, the free tier is notably generous: up to 60 model requests per minute and ~1,000 requests per day, with access to Gemini 2.5 Pro and its 1M‑token context. In hindsight, that surpasses my current Claude Pro limits—which says a lot.

Setting up Gemini CLI

Ensure you have Node.js version 20 or above installed.

Execute the following command in your terminal:

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

Or install with:

npm install -g @google/gemini-cli

Then, run the CLI from anywhere in your terminal (ideally in your project directory):

gemini

You should see the following upon running the command above:

Select "Login with Google". Now you have a terminal CLI you can code with your favourite IDE, for FREE.

But say you do hit the limit:

1. Head to aistudio.google.com and click on the 'Get API key' button.
2. Create a key and add it to your environment files:

GEMINI_API_KEY="KEY-HERE"

Now go back to Gemini CLI, run /auth and select "Use Gemini API Key". With this, you can switch between Google login and API key auth daily for longer usage of Gemini CLI.

Supercharging Gemini CLI capability with Roo Code

But wait! You could do so much than just using Gemini CLI directly in your terminal. For instance, using together with an open source AI-powered autonomous coding agent integrated in VS Code: Roo Code

1. Head over to "Extensions" in your VS Code and search for Roo Code.
2. Click on the small settings icon near Roo Code, and select "Install Specific Version...".
3. Select "3.22.4" Version
4. Once installed, open Roo Code and head over to its settings.
5. Select "Gemini CLI" in your API Provider. Ensure you have authenticated with Gemini CLI before running Roo Code.

All done! Roo Code has several modes you can choose from: Code, Architect, Ask, Debug, and my favourite, Orchestrator. Orchestrator allows the LLM to coordinate tasks across multiple modes, breaking complex problems into multiple subtasks, and more.

Heads‑up: Roo Code’s prompts can be quite long, so token usage climbs with large refactors or architecture migrations. That said, for everyday tasks, the free Gemini is usually more than enough.

Closing thoughts

While paid AI coding agents are scarily great in coding nowadays, an overreliance on said agents would result in paying for higher-tier plans just to maintain the same limit. IMO, great for rapid prototype, shipping ideas to reality, but avoid using it for production cases.

Happy Building!