Show HN: ThinkReview open source browser Copilot GitLab and ADO PRs(Ollama)
github.comOver the last few months I’ve been building a lightweight, open-source “copilot” that runs directly in your browser and helps you review Pull Requests / Merge Requests without sending code to any external service unless you choose to.
ThinkReview just gained two major updates:
1. *It is now fully open source* 2. *It now supports Ollama, so you can run LLMs locally for private code reviews*
---
## What is ThinkReview?
ThinkReview is a browser extension that attaches itself to the native PR/MR UI in:
- GitLab (self-hosted or SaaS) - Azure DevOps - GitHub - Bitbucket
Instead of acting like CodeRabbit, CodeAnt, or CI-connected bots that auto-comment on your PRs, ThinkReview does something different:
- It gives you a *private chat window* attached to the diff view. - You can ask the model questions about the MR, explore logic, identify potential bugs, or generate draft comments. - The AI doesn’t post anything automatically — you stay in control.
This works well for developers who still do most of their reviewing in the browser and don’t want a noisy bot writing public comments.
---
## Demo (GIF)
<img src="https://firebasestorage.googleapis.com/v0/b/thinkgpt.firebas..." width="600" />
---
## Why open source?
A lot of early users (especially from companies running self-hosted GitLab) asked for:
- transparency around where code goes - the ability to self-audit - control over model choice - contributions and community fixes
Repo: https://github.com/Thinkode/thinkreview-browser-extension
It’s built with standard browser APIs and a small LLM integration layer, so it’s very hackable.
---
## Ollama Support (local LLMs)
As of v1.4.0, you can point ThinkReview to your local Ollama instance. This lets you run any supported model:
- Qwen Coder - Llama 3 - DeepSeek - Codestral - Any other Ollama model
### Why this matters
- Your code *never leaves your machine* - Zero cost - Works cleanly with self-hosted GitLab / air-gapped setups - No API keys, no vendor lock-in - You can swap models instantly
If you prefer speed, you can still use cloud LLMs; if you prefer privacy, Ollama works surprisingly well (tested ~50s on a Mac Mini M4).
---
## Installation
Works on any Chromium browser:
Chrome Web Store: https://chromewebstore.google.com/detail/thinkreview-ai-code...
No backend server required. Configuration is minimal.
---
## Looking for feedback
HN has a lot of people who:
- review PRs daily - care about developer ergonomics - run self-hosted GitLab - have thoughts about local vs cloud LLM workflows - like hacking on browser extensions
If you try it, I’d love feedback on:
- UI/UX improvements - additional provider integrations - other platforms to support - performance / caching ideas for local LLMs
Discussions / issues: https://github.com/Thinkode/thinkreview-browser-extension or https://thinkreview.dev/contact
Thanks for reading, *Jay*
No comments yet.