AI coding tools have made writing code much faster.
But we noticed a new bottleneck in our development workflow:
Code review.
When developers can create code faster, more PRs are opened. But the number of reviewers doesn't increase at the same speed.
Why review is the bottleneck
This is what we started seeing in our workflow:
The problem wasn't that our reviewers were slow.
There was simply too much code for humans to review manually.
So we started experimenting with AI for the first pass of PR review.
We built Open PR
We built Open PR, an open-source AI agent for reviewing GitHub pull requests.
GitHub — TOMOSIA-VIETNAM/open-pr
The idea is simple:
Developer → Pull Request → AI Review → Human Review → Merge
AI handles the initial pass, while humans make the final decisions.
We're particularly interested in problems that go beyond formatting or simple code smells — things like business logic, unexpected side effects, missing test cases, and inconsistencies with the existing codebase.
Setup with Claude Code
First, make sure you have GitHub CLI installed and authenticated:
gh auth login
Then add the Open PR plugin marketplace and install the plugin:
/plugin marketplace add TOMOSIA-VIETNAM/open-pr
/plugin install open-pr@open-pr
Once installed, you can review a GitHub PR directly from Claude Code:
/open-pr:review <PR_URL>
For example:
/open-pr:review https://github.com/your-org/your-repo/pull/123
Claude Code will analyze the PR and return review findings that you can inspect and discuss before merging.
Still an experiment
We don't think AI code review is solved.
AI can miss important issues, misunderstand business context, or produce false positives. That's why we're treating Open PR as an assistant for reviewers, not a replacement for them.
We're sharing it openly because we'd like to learn from other teams.
Has AI made your PR review process the new bottleneck?
If you're experimenting with AI-assisted development, we'd love to hear what your workflow looks like.
Open PR on GitHub