AI Code Review sidebar on Azure Devops / TFS PRs

[mshenawy22]

🏷️ Discussion Type

Question

Body

I would like to share with you the tool that I have been building/bootstrapping for the last 8 months
I built it initially to fix my own problem of reviewing PRs for my team , then it kept on expanding

It's called ThinkReview opensource on Github

And it works out of the box without setup on Azure Devops once you install the browser extension
It also supports TFS or azure server with custom domain (you just need to add the custom domain in the extension and your PAT) takes less than 30 seconds

As you see in the picture if providers scoring for code quality , secuity and best practices. Gives you code suggestions for each of the points.

You could also create custom agents with custom prompts and section and it would appear as a a new tab.

it does not auto post comments like other tools , the approach here is that ot empowers the manual review , the reviewer would have to post the comment manually at the end under their own name .

The tool works by default on the diff of the PR , but it can also extend to have full context on the whole repo if you gave it permissions to do so.

The tool is powered by a selection of frontier and opensource models you could choose from . in addition to an integration with Ollama if your machine is powerful enough

I hope the community finds this as a useful tool and I'd like to hear any suggestions or questions

the extension is available on both chrome and firefox stores : https://thinkreview.dev/

[Ganesh-Code-Lives]

Interesting idea, but here’s the honest take:

• The “manual post” approach is good for control, but it also slows teams down — most people prefer semi-automated comments.
• Accuracy matters more than features. If suggestions aren’t consistently useful, devs will ignore it.
• “Works out of the box” is strong — but you should clearly show how it compares vs existing tools (Copilot, CodeRabbit, etc.).

Main suggestion:
Focus on signal quality + real-world benchmarks, not just features. That’s what will make people adopt it.

Overall, solid direction — but adoption will depend on how reliable the insights actually are.

[Ganesh-Code-Lives]

if you like it , mark it as answered

[mshenawy22]

Explain what are "semi automated comments" and how they different than full automated comments ??On 4 May 2026 09:34, Ganesh Mamidisethy ***@***.***> wrote: Interesting idea, but here’s the honest take: The “manual post” approach is good for control, but it also slows teams down — most people prefer semi-automated comments.Accuracy matters more than features. If suggestions aren’t consistently useful, devs will ignore it.“Works out of the box” is strong — but you should clearly show how it compares vs existing tools (Copilot, CodeRabbit, etc.). Main suggestion: Focus on signal quality + real-world benchmarks, not just features. That’s what will make people adopt it. Overall, solid direction — but adoption will depend on how reliable the insights actually are. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[Ganesh-Code-Lives]

Semi-automated comments = the tool suggests feedback, but the reviewer checks and posts it manually.

Fully automated comments = the tool posts comments directly on the PR without human approval.

In short:
Semi-auto = controlled, fewer mistakes
Full-auto = faster, but can spam or be wrong

[mshenawy22]

Thank you very much for your feedback On 4 May 2026 09:34, Ganesh Mamidisethy ***@***.***> wrote: Interesting idea, but here’s the honest take: The “manual post” approach is good for control, but it also slows teams down — most people prefer semi-automated comments.Accuracy matters more than features. If suggestions aren’t consistently useful, devs will ignore it.“Works out of the box” is strong — but you should clearly show how it compares vs existing tools (Copilot, CodeRabbit, etc.). Main suggestion: Focus on signal quality + real-world benchmarks, not just features. That’s what will make people adopt it. Overall, solid direction — but adoption will depend on how reliable the insights actually are. —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[Ganesh-Code-Lives]

Thanks for the response.

One suggestion — if you want better adoption, consider adding a semi-automated mode (review → approve → post). It keeps control while reducing friction.

Also, if possible, share some real-world benchmarks or examples comparing your tool with others — that would make the value much clearer.

Overall, solid idea — execution and accuracy will decide how far it goes.

[mshenawy22]

@Ganesh-Code-Lives First off, I really love your notes here—thank you so much for taking the time to write this out. Your perspective on the approval buttons to auto comment , is incredibly helpful.

Since you clearly have a great eye for this space, I would absolutely love to get your honest feedback on ThinkReview. If you have the bandwidth to give it a quick spin, I’d be thrilled to upgrade your account to a Professional subscription for free as a thank-you for your time and insights.

Let me know if you'd be open to trying it out! Also, I'd love to connect with you—feel free to find me on https://x.com/ThinkodeAI f you're ever open to chatting."