Hello AI Developers! ![]()
Code reviews are vital, but let’s be honest—staring at massive diffs for hours is exhausting, and it is easy for security vulnerabilities or edge-case logic flaws to slip through when fatigue sets in.
To help solve this, I wanted to share a project I’ve been working on: ThinkReview. It is a production-ready, open-source browser extension (Manifest V3) that natively injects AI-powered code reviews and an interactive copilot directly into your pull/merge request pages.
What Makes It Different?
Unlike traditional tools that require messy CI/CD pipelines, repository admin access, or complex GitHub Actions setup, ThinkReview operates entirely within the browser. It automatically detects when you are viewing a PR/MR and pulls the diff context in real-time.
**
Here is a breakdown of its core features**:
Zero-Setup Agentic Reviews: Instantly scores your PRs and highlights critical bugs or security risks right inside a collapsible side-panel on your Git platform.
Custom Review Agents: You can configure up to 10 distinct custom agents tailored to your team’s architecture (e.g., a Senior Pen-Tester agent or a Full-Stack Performance Lead agent).
Bring Your Own Key (BYOK): Out of the box, it connects directly to OpenRouter, allowing you to choose from over 15+ frontier models (including Gemini 1.5 Pro/Flash, Claude, or GPT) with zero data stored on our end.
Private & Local AI: For enterprise teams or sensitive repositories, it features complete offline support via Ollama. Your code never leaves your local machine.
Interactive Copilot: You can chat directly with your Git diff using natural language to ask for alternative implementations, generate structured reviewer comments, or push suggestions directly into your IDE with one click.
Architecture & Tech Stack
The extension is designed to be lightweight, secure, and respectful of developer privacy:
Frontend: Built as a WebExtension compatible with Chrome, Edge, Brave, and Firefox.
Security: Uses Google OAuth for secure authentication. No API keys are hardcoded; OpenRouter requests are called directly from your browser runtime to openrouter.ai/api/v1.
Data Privacy: Code diffs exist only in memory during the analysis cycle and are never used for AI training models.
Open Source & Contributing
We have recently gone fully open source under the GNU Affero General Public License v3.0 (AGPL-3.0) to foster transparency, trust, and community collaboration.
We would love for the Google AI community to audit the code, try it out during your daily workflows, and help us extend its functionality.
Official Website: thinkreview.dev
GitHub Repository: Thinkode/thinkreview-browser-extension
Feedback Wanted!
[1]
Footnotes ↩︎