Kilo Code Reviewer - Automatic AI-powered code reviews the moment you open a PR
Automated code review agents that analyze pull requests, suggest improvements, catch bugs, and ensure code quality standards. Pick from 500+ models (Claude, GPT, Gemini, and several free options) to get instant feedback before merging.
Replies
JDoodle.ai
Very interesting product. Out of supported 500+ models, is there any recommended coding models list you guys suggest, as only few of them good in coding and review.
Humans in the Loop
@gokuljd thanks for the support, Gokul!
Good question. FWIW There's an on-going poll/thread on the topic here. Suggestions include @Claude by Anthropic's Sonnet 4.5 and Opus 4.5 (72%), @Gemini 3 (13%), @OpenAI's GPT-5.2 (12%).
How about you? Any preferences?
Humans in the Loop
@gokuljd Quick update: FYI @Kilo Code maintains a leaderboard with the top coding models.
Current ranking, based on real-world testing:
@Claude by Anthropic Opus 4.5
@Grok Code Fast 1
@OpenAI GPT-5.2
Hope it helps! Source: kilo.ai/leaderboard
Triforce Todos
If this reduces review back-and-forth even 30%, that’s already a win. Wishing you all the best :)
Humans in the Loop
exactly! help us spread the word on X
file.coffee
This is wonderful. We've been using this Code Review mode since early testing and given tonnes of tips/feedback, love where it's at today!
Humans in the Loop
awesome! what do you enjoy the most about this feature?
file.coffee
@fmerian just that it's flexible. we pay for 1 tool, which does everything.
Humans in the Loop
absolutely lovin' it! @Kilo Code, the ultimate all-in-one agentic engineering platform? spread the word on X!
file.coffee
@fmerian yep! it's great!
Product Hunt
@curiouskitty We designed it to be non-invasive with easy setup that can be configured deeper if wanted.
Basically, you just turn it on in the Kilo dashboard, and reviews will start happening automatically. You can also adjust the review strictness (reducing reviews to only serious issues), and pick from specific focus areas.
Furthermore, you can provide custom instructions to steer the AI in a direction that makes sense for your team!
If you are a solo-founder or are vibe coding, having this code review gives you another check on the quality and security of your code. This is far better than skipping them and is faster & less expensive than hiring a fractional worker to do it.
Humans in the Loop
You're spot on, Josh.
Vibe coding apps keep hitting vulnerabilities: exposing secrets, access misconfigurations, hardcoded credentials. Hopefully @Kilo Code could help makers build products more secure, faster.
@kamil_maksymowicz There are a few ways to access Code Reviewer.
1. In our VS Code extension, by just selecting "Review" mode in the mode selector dropdown. This fits more into traditional coding workflows that occur in the IDE.
2. At app.kilo.ai/code-reviews, where you connect your GitHub repo once, and Code Reviews will appear on every PR. This is more of a Grammarly-type experience, in that you really only have to check in from the browser, if desired.
Kilo Code
Literally yesterday, Kilo Code Reviewer saved me from pushing a silly but hard-to-notice bug to our production docs system!
Humans in the Loop
@realolearycrew awesome! any chance you could share the link to the PR? would love to have real-world examples
Does Kilo limit its analysis strictly to the diff / selected changes, or can it reason about a wider scope (files, modules, usage elsewhere) in a way that adapts to the size of the repository?
From my recent experiments with AI code reviewers, most of them either:
stay diff-only, or
try to scan “everything” and quickly become impractical on larger repos due to long processing times.
What I’m missing is something more adaptive — e.g. scope expansion that’s proportional to repo size or dependency graph, rather than a fixed “deep analysis” mode. Curious how Kilo handles this today, and whether repo-aware scoping is part of the design.
Tried it today to review some edge functions in Supabase before deploying, and that wasn't too obscure and caught some things. Excellent feature!
Humans in the Loop
Thank you! Curious what model(s) did you use here?
@fmerian To start I used Sonnet 4.5 because it's a known and trusted entity, but I'm experimenting with some free ones to see how they compare.
Humans in the Loop
please keep us posted on what's good. there's an on-going poll/thread here: https://producthunt.com/p/claude/what-s-the-best-ai-model-for-coding
looking forward to your thoughts
I played with codereview from the Kilocode web interface, and it seems to only allow one configuration done.
I can't seem to find a way to have different reviews for different repos with different options/models/prompts.
However saying that, this has great potential to be an amazing addition to my workflows for PR's once developed a bit more. Update: PR's in GitHub, while only 1 config/prompt overall at any time, work very good in GH and fairly fast and seamless when it doesn't stop with a status of "interrupted". Eh, I take back the fairly fast. I have some commits that take over 30 min to do a code review on a basic readme file edit, and others that just hang forever in the Kilo web interface and never process. I have to retry quite a lot. It's nice to have implemented, but I am finding my workflow greatly slowed down just waiting for Kilo to review if it ever does. I' played with different models, but it is intermittent and makes the product less usable and more of a lets see what it says if we wait. What was created as a bottleneck reducer became the bottleneck.
Using it inside the KiloCode extension seems better for a lot of my current use cases while actively developing.
I like that I can switch to review mode and have it review uncommited changes before I commit.
While it does say "Recommendation: APPROVE" in the main response, I especially like that the very last bit of the summary is "APPROVE" so I can have faster workflow and decide to read more otherwise.
It still feels very early in implementation, but I'm liking what I see so far.