AI Reviews - Instant AI code review.
Graphite Chat finds critical bugs and security issues before they reach production.
Instant reviews
1-click fixes
High signal
Custom rules
Graphite Chat
342 issues found
Potential edge case
Documentation issue
Code quality/style
Performance issue
Accidentally committed code
Security issue
Logic bug
1482
PRs reviewed
281
Issues led to fixes
3.5%
Comments downvoted
const port = 3000;
Unresolved thread: line +3
Logic bug
Change the port from 3000 to 8000 to match the exported port in the tests. The CI test is failing because it expects the server to run on port 8000.
Suggested change
// Change this line
const port = 3000;
// To this:
const port = 8000;
Commit suggestion
Filters Define information to consider when determining types of comments not to leave.
- Dialogs do not need explicit onClose handlers.
- Don't leave comments about naming changes that could be considered nits.
Commit suggestion
Update server.js
Commit suggestion
Commit changes
Custom rules
Define rules to detect issues in your codebase in addition to general bugs.
- Implement rate limiting and request validation to prevent abuse.
- Verify the integrity of third-party libraries and dependencies before use.
Get feedback on every PR in seconds, not hours.
Testimonial
It's different from other AI tools because...it actually works.Brian Michel, Software Engineer
Getting started
How it works
Start catching bugs in under 3 minutes.
Connect GitHub
One-click authorization. No config files. No setup.
- Your email
- Sign up with GitHub
Get instant reviews
Diamond analyzes every pull request in seconds, flagging only real issues.
Graphite left comments
Fix with one click
Accept AI-powered suggestions or customize the fix. Ship with confidence.
Features
Why teams choose Graphite Chat
The features that help teams review faster, catch more bugs, and enforce best practices.
More impact, less noise
Teams that use Graphite Chat catch more critical issues with fewer false positives than competing solutions, from benchmark tests.
Less than 5% negative comment rate
Instant reviews
Graphite Chat integrates directly with GitHub where you can take suggestions in 1-click. No context switching needed.
Seamless integration with GitHub
Custom rules
Scale your team’s best practices. Write your team’s preferred rules in plain language, or choose from templates. Graphite Chat enforces them across every PR, instantly.
Adhere to your team's best practices
- OWASP Security
- Airbnb JavaScript style guide
- a11y Accessibility
- PEP Python style guide
- Google Java style guide
- Google TypeScript style guide
- Airbnb Ruby style guide
- Google Go style guide
Logic bug
The condition in the if statement should be inverted. Change it to:
if (!Object.keys(ASYNC_JOB_REGISTRY).includes(kind)) {
Potential edge case
readlink -f is not available on macOS by default. A more portable solution would be:
filepath=$(perl -e 'use Cwd "abs_path";print abs_path(shift)' "$0")
This achieves the same result but works across both Linux and macOS environments.
Security issue
Please remove this token logging. Exposing GitHub access tokens in logs creates a security risk, as these tokens grant API access and should be treated as sensitive credentials.
Accidentally committed code
Adding || true to this condition causes the function to unconditionally return "BINARY", bypassing the binary detection logic. This appears to be unintentional and should be removed to restore the original file type detection behavior.
Performance
The await inside Promise.all([...]) is preventing parallel execution of these promises. Since Promise.all expects an array of promises, the await here will cause the stack dependency check to complete before the merge job query begins. To achieve the intended parallel execution, remove the await from db.stack.getStackDependencyForPr().
Code style/quality
The callsite value markNotifAsRead doesn't match the function name markAsReadEndpoint.
Documentation issue
The description 'Get auth tokens for user' doesn't match the command's actual functionality. Consider changing it to 'Run async job queue and send a test email'.
Use cases
Catch more than just typos
- Logic bug
- Potential edge case
- Security issue
- Accidentally committed code
- Performance
- Code style/quality
- Documentation issue
Testimonials
What customers are saying
See what other engineers have said about AI Reviews.
Konrad Komorowski, Software Engineer
Graphite Chat is very high signal, unlike other tools on the market. It found a months old bug in our codebase, that no human engineer could spot and so far, every highlighted issue has been a real problem.
Brian Michel, Software Engineer
It's different from other AI tools because... it actually works. It doesn't blindly say 'hey here's a typo,' it gives well reasoned recommendations based on the context of what I'm trying to build and what I'm trying to build it in.
Mike Deeks, Infra Engineer
Graphite Chat is currently my favorite thing. No fuss, easy to set up, easy to use, super helpful. I inserted our credit card and our lives got better.
Saket Gejji, CTO
We're seeing it comment on nits that we definitely would have missed, which is fantastic for improving our code quality.
Ruven Chu, Software Engineer
Graphite Chat has been great for us! It does a fantastic job at finding real problems with our code and not delivering a lot of noise.
Ali Andaç, Software Engineer
Graphite Chat is a handy teammate who checks the PR quickly and saves your hours before merging your change. With this instant boost, development process time could decrease dramatically and ensure code quality level. Suggesting teams who aim for speed.
Reggie Escobar, Software Engineer
It's really fast (it caught things before our CI did). It caught things before PR reviewers did.
Adam Walker, Co-Founder & CEO
I've been using it for two weeks now and have to say, while it has only triggered a handful of times, each time has been an incredibly helpful insight on something I overlooked as well as other members of my team.
FAQ
How do I enable AI code reviews?
Still have questions? Reach out to our team.