I recently built Deni AI Harbor to fix a pain point many OSS maintainers hit during PR review:
- repetitive comment noise,
- stale approvals after PR updates,
- unclear check/status feedback during reruns.
Deni AI Harbor is designed to run on GitHub PR workflows and help keep review flow predictable.
What it does
- Reviews PRs on lifecycle events (
opened,reopened,synchronized, etc.) - Supports manual mention-based review triggers
- Posts inline suggestions and summary review comments
- Shows progress and final state through a dedicated status/check path
- Clears bot approvals before rerun to avoid stale approval states
- Returns clearer completion outcomes in CI/check visibility
Why I built it
As repositories get more active, iteration speed increases but review coordination gets messy. I wanted a tool that feels transparent:
- deterministic behavior,
- explicit status reporting,
- low-surprise automation.
Current status
It’s in active development and now open to real-world use.
If you try it, I'd love feedback on:
- false positives / noisy suggestions
- how it behaves under frequent reruns
- anything odd in approval or status behavior
Links
GitHub: https://github.com/deniai-app/harbor
Install: https://github.com/apps/deni-ai-harbor
Top comments (0)