This is a submission for the GitHub Copilot CLI Challenge

AI Code Reviews for $0.00? I Built a Provider-Agnostic Reviewer with GitHub Copilot CLI

What if AI code reviews didn’t cost $50/month—or lock you into a single model?


I ran into this problem while maintaining side projects, so I built an AI Code Reviewer that works with any LLM provider and can cost literally $0 per PR.

What I Built

AI Code Reviewer is an open-source GitHub Action that performs automated, AI-powered code reviews on every pull request.

In short:

• 🔌 Works with any OpenAI-compatible API (Groq, DeepSeek, Mistral, Gemini, Anthropic, OpenAI, etc.)


• 🧠 Uses a custom compression format (TOON encoding) to reduce token usage by 50–70%
  


• 💸 Supports free LLM providers → $0.00 reviews


• ⚡ Setup time: ~30 seconds


• 🔓 Fully open source (MIT)

This wasn’t just about building another AI tool—it was about giving developers back control over cost, models, and infrastructure.

Demo

📸 Github PR for sample:

“https://github.com/tusharshah21/ai-code-reviewer/pull/2”

📸 Screenshot:

“AI Code Reviewer leaving structured review comments on a GitHub PR”

Why I Built This

Most AI review tools today share the same problems:

• Vendor lock-in


• Opaque pricing


• Closed source


• Expensive for students, indie hackers, and OSS maintainers

I wanted the opposite:

You choose the model. You control the cost. You own the workflow.

My Experience with GitHub Copilot CLI (The Real Game-Changer)

I’m primarily a backend developer.


GitHub Actions, Octokit, and CI ergonomics were new territory for me.

Instead of tab-hopping between docs, blog posts, and Stack Overflow, I stayed inside my terminal and used GitHub Copilot CLI as a learning companion.

Concrete Examples

1. Fetching PR diffs with Octokit

bash

gh copilot suggest "how do i fetch pr diff using octokit"