Curious about your code.
AI code review for GitHub, powered by the model you choose. It looks for bugs, double-checks what it finds, and approves the rest.
feat: let customers cancel orders #42
Open anna wants to merge 3 commits into main
Hansi requested changes Changes requested
B Needs changes before merging
Adds order cancellation. One bug lets shipped orders be cancelled; everything else looks good.
Missing await
find() returns a promise, so order.status is always undefined and shipped
orders can be cancelled.
Comments on bugs, not style
Hansi looks for the bugs you would want a teammate to catch: a missing await, an inverted check, an off-by-one. A second pass drops anything it can't confirm in the code.
3 possible issues found
2 dropped after double-checking
1 comment posted
Approves like a teammate
Clean pull requests get approved. When something is wrong, Hansi requests changes, and once you push the fix it checks again and approves.
Hansi requested changes
anna pushed 1 commit
Hansi approved these changes
A grade for every pull request
Each review ends with a grade from S, ready to merge, to F, do not merge. Open findings cap the grade, so it always matches what Hansi found.
Bring your own model
Hansi runs on your key and the model you pick, so how deep the reviews go and what they cost is your call. You can review with a stronger model and double-check with a cheaper one.
- OpenAI
- Anthropic
- Amazon Bedrock
- Google Gemini
- xAI
- OpenRouter
- Cerebras
Free during the beta
Unlimited repositories and reviews while Hansi is in beta. You pay your model provider directly for the tokens each review uses, and Hansi shows you the cost of every review.
$0per month
- Unlimited repositories
- Reviews on every push
- Approvals, grades, and inline fixes
- Your own model and API key
Questions
Which models can Hansi use?
Any model from OpenAI, Anthropic, xAI, Google, OpenRouter, or Cerebras, plus any OpenAI-compatible endpoint. You can use a stronger model to review and a cheaper one to double-check findings.
How good are the reviews?
That depends mostly on the model you choose. Every model gets the same tools and the same skeptical second pass, but a stronger model will usually spot subtler bugs. Every review shows its cost, so you can weigh one against the other.
What does it cost?
Hansi is free during the beta. You pay your model provider directly for the tokens each review uses, and every review shows its token usage and cost.
What happens to my code?
Hansi reads the pull request to review it and sends the relevant code to the model provider you chose. The checkout is deleted when the review finishes.
Will it flood my pull requests with comments?
Hansi is built to comment only on bugs, and a second pass drops anything it cannot confirm in the code. Clean pull requests usually get no comments, just an approval.
Can it approve pull requests?
Yes. Hansi approves clean pull requests and requests changes when it finds a problem it could confirm. It never approves pull requests from people without write access to the repository.