open-source browser extension: AI code review copilot for Bitbucket cloud PRs

jay shenawy

Hi everyone,

I built an open-source browser extension called ThinkReview that adds AI-powered code reviews directly on Bitbucket pull request pages, so you can get AI  analysis without leaving the PR and without wiring up heavy integrations.

Source code

• GitHub: https://github.com/Thinkode/thinkreview-browser-extension
• Chrome Web Store: https://chromewebstore.google.com/detail/thinkreview-ai-code-revie/bpgkhgbchmlmpjjpmlaiejhnnbkdjdjn
• Firefox Add-ons: https://addons.mozilla.org/en-GB/firefox/addon/thinkreview-code-review/

 

 

What it does

• Detects when you’re on a Bitbucket PR and shows a review panel in the page
• Summarizes changes, highlights risks, and suggests improvements (depending on the model you use)
• Includes a chat-style flow so you can chat with your PR
• Supports the latest frontier models in the cloud, plus local models via Ollama if you want reviews to stay on your machine
• you can add up to 10 custom agents to review your PR

Bitbucket setup (60 seconds)

• Bitbucket Cloud (bitbucket.org): install the extension
• You will need to generate an API token with scopes (read:repository:bitbucket , read:pullrequestBitbuket)
• Add the api token in the extension (credentials are saved locally in the extension local storage)
• and open a PR — it should work with your normal browser session.
• Bitbucket Data Center / Server / custom domain: add your instance hostname in the extension settings (similar idea to allowlisting a self-hosted domain), then open a PR on that host.
  Full step by step :https://thinkreview.dev/docs/bitbucket-integration

If you try it on Bitbucket cloud , I’d love feedback on what works well (and what doesn’t) in real workflows.

Thanks for reading.

Nick Sudo

This looks like a useful approach to code review, especially for teams that deal with a lot of pull requests. Having AI review the changes directly in Bitbucket could save developers some time and help catch issues that might otherwise be missed.

I also like that it supports local models. For some teams, keeping source code on their own machines can be an important consideration when using AI tools.

That said, I would still treat AI feedback as a second set of eyes rather than a replacement for a developer. Context matters a lot in code reviews, and an AI suggestion isn't always going to understand the full purpose behind a change.

As AI Development becomes more common in software teams, I think tools like this will be most useful when they support developers without taking away the human review process. AppliconSoft is also working in this broader development space, so this is an interesting direction to follow.