In 2021, I was deep in Salesforce development, reviewing pull requests, fixing edge cases, and trying to ship clean code. AI code reviews weren’t a thing back then.
A few years ago, code reviews meant reading every line closely, juggling context across files, bugging other developers for explainers, and dropping comments that often went unnoticed.
Then in 2025, I joined Bito.ai. It’s been more than 5 months, and I’ve been using Bito’s AI Code Review Agent on real pull requests.
Why I chose to work with Bito
I came across Amar Goel, Bito’s cofounder and CEO, while I was working in technical writing and developer marketing. We started talking, and he shared what they were building. An AI agent that reviews pull requests in GitHub without storing your code.
That felt so cool. I kept thinking about 2021.
Back then, every pull request I opened meant asking teammates to review logic, naming, structure, and edge cases. It took time, added friction, and sometimes, things slipped through.
The idea of getting instant, contextual suggestions inside a PR, without giving up your code, felt like something I would have wanted as a developer. It made sense, felt practical. It felt like something made for how developers actually work.
I joined Bito shortly after.
This post is the result of 150 days working with Bito and using the AI Code Review Agent on my own code. Every example I share here comes from that experience.
AI in pull requests: What changed for me?
Back in 2021, most of my pull request reviews used to follow the same pattern. I would open the diff, scroll through my changes, and try to catch anything that felt off.
Sometimes I would miss things. Sometimes I would forget what I was thinking when I wrote the logic in the first place.
Once I started using Bito’s AI Code Review Agent, I noticed the difference right away.
The AI-generated comments showed up directly in the pull request.
They pointed out specific lines and explained why something could be improved.
The suggestions were clear. If a function was too long or a condition could be simplified, the agent said so.
If I reused a pattern that could be abstracted, it highlighted that too.
I did not need to change how I reviewed pull requests. I just had more context in the same place. That saved time and made the feedback loop tighter.
The inline suggestions inside my PRs were the easiest part of the experience to adopt. I wrote more about that experience in this post on how I started chatting with my AI code reviewer.
Next, I’ll walk through what Bito is doing now, what new features we are working on, and why I’m betting on it.
AI code reviews: What we’re doing at Bito?
Over the past 150 days, Bito kept adding features that made pull request reviews smoother and more aligned with real-world development. And I tried them all.
We’ve already listed out all the major features that Bito’s AI Code Review Agent offers in the product page. Here’s a quick overview:
- Reviews pull requests
- Gives inline, contextual suggestions
- Catches code smells, logic issues, and style problems
- Learns from your feedback and adapts
- Doesn’t store your code or use it for training … and much more.
In this section, I specifically want to talk about the major updates Bito released and things I did, since I joined:
1/ Custom code review rules
This is one of the latest and coolest updates Bito dropped after I joined. Bito’s AI Code Review Agent lets you enforce your own coding standards directly in PRs:
Automatic learning from feedback: When I mark a suggestion as irrelevant, the AI learns not to offer that again. Once you mark a similar suggestion as irrelevant three time, Bito creates a custom rule. Read about it here.
Custom rule uploads: This is the latest update! You can now create a custom guideline directly within your Bito dashboard. I created a video walkthrough. Watch it on YouTube here.
More on this is detailed in How I Personalized Bito’s AI Code Review Suggestions, which is actually about defining rules. The feature works as expected!
2/ Multi-product seat based billing
Earlier in June, Bito also rolled out a redesigned dashboard for managing users and billing. This was a big deal for teams.
The new member management view shows exactly how many seats your workspace has, how they’re distributed between IDE users and pull request reviewers, and who’s using what.
It’s now seat-based by product, which makes scaling much easier to track.
There’s also an auto-assignment option. Meaning: new dev joins and they get a seat automatically. Or turn it off and do it manually. That’s up to the admin.
I also created a video walkthrough to explain this update. Watch it on YouTube here.
3/ Chatting with the AI Code Review Agent
This feature was launched in April. It gives you the ability to chat directly with Bito’s AI Code Review Agent. You can ask the agent follow-up questions on its suggestions. Things like:
- “Why is this a problem?”
- “Can you suggest a cleaner way to do this?”
- “What’s an alternative approach?”
I tried this in one of my own PRs and wrote about it in Chatting With My AI Code Review Agent.
Also worth noting: Bito supports over 20 human languages in this chat experience. I mostly use English, but I tested a few queries in Hindi just to see. Works just fine.
4/ Agentic code reviews
This was a game-changer. Bito’s AI Code Review Agent is now fully agentic. That means it no longer follows a fixed chain-of-thought pipeline.
Instead, it dynamically figures out what context matters, explores the code more freely, and generates suggestions based on real structure and patterns.
5/ Multilingual code reviews
Bito now supports over 20 human languages in review comments. That includes English, Hindi, Chinese, and Spanish.
If you’re working with global teams or reviewing code with non-English speakers, this is just one less thing to worry about.
Read the doc for the steps.
6/ Amazon Nova Lite 1.0 in Bito
With the release of Bito’s free tier, Nova Lite became a key model powering the Code Review Agent for individual developers.
What that means for developers? You get free AI code reviews for everyday tasks that don’t need deep reasoning. The experience still feels tight and helpful.
Amazon even featured this in a case study with Bito. You can read it here: Amazon Nova + Bito Case Study
Other Updates
Bito rolls out updates weekly. You can find all release notes and documentation on our official site.
Most of the features I used during these 120 days came from regular product iterations that improved my workflow quietly in the background.
See the full release changelog and docs here: Bito documentation.
Bito, its competitors, and benchmarking
If you’re a marketer in tech, you can’t just skim docs and pitch jargons. You need to understand the product deeply. And that means using it.
For me, that also meant trying out the competition. Because if I’m going to talk about Bito, I need to know exactly how it stacks up.
So I did what any curious developer-marketer would do. I opened real pull requests and used Bito side by side with the other tools. A few examples:
Bito vs Coderabbit
This one was a full comparison. I took the same PR and ran it through both tools. Bito gave sharper, more relevant suggestions. It caught things I actually cared about.
Coderabbit left more noise than value. Less signal, more cleanup. I documented everything here: Bito vs Coderabbit
Bito vs GitHub Copilot
Then came Copilot. It’s great for writing code in the IDE, sure. But when it comes to reviewing code in pull requests, it’s just not built for that. No inline feedback, no context, no real review flow.
Bito, on the other hand, lives inside your PR. It gives you comments where they matter, grounded in the actual diff. You can see my detailed comparison here: How is Bito Different from GitHub Copilot?
Benchmarking AI code review tools
After using Bito for a few months, I also wanted to see how it stacked up in more structured comparisons. The team at Bito has actually built a proper benchmarking setup for this.
We test the tool on a known set of code issues across multiple languages, and compare it with other tools in the space. Based on the benchmarking:
Bito performed well. It had the highest issue coverage and consistently caught more high-severity bugs in languages like TypeScript, Python, JavaScript, Go, and Java. That kind of data gave me more confidence in what I was seeing in my own pull requests.
You can check out the whole thing here: Benchmarking the Best AI Code Review Tool
Final thoughts
150 days ago, I joined Bito to understand how AI fits into real code reviews. Since then, I have used Bito on live PRs, tried every feature it shipped, and compared it directly with other tools.
As someone in developer marketing, I believe in understanding the product deeply. That means using it the way real developers do.
After five months, I see why Bito is different. It provides codebase aware PR suggestions, keeps your code private, respects your coding standards, and makes PR reviews more efficient.
Security matters. Many AI (or non-AI) tools still store your code or use it to train their models. That should not happen in a secure code review process. I wrote about that here: Secure Code Review Process
If you are curious, try Bito on a real pull request. That is the only way to see what it can actually do:
Top comments (5)
Very helpful. Thanks for sharing!
I've enjoyed all of the research you've put into this project, it adds up and you can see how much you care about the details
Thank you, Nathan. I see you popping up in my posts here a lot! Thank you again! Would love to know how you came across me here.
Hey Sushrut, I try to be very active on DEV, and since I write, I appreciate it when folks leave comments on my articles.
Thoughtful read and I appreciate the personal perspective here. Not enough people share their ins and outs, ups and downs like this. Thanks for sharing yours!