Hey HN! I'm Hussam, a Senior Algorithm Engineer at CodiumAI. We launched PR Agent a little over a year ago — it’s a free and open-source tool that helps with reviewing pull requests. The response has been great and more than 70 contributors have added to the project.
Today, we're launching PR Agent as a browser extension.
Key features:
- AI Chat in PRs: You can ask questions about specific code changes and the context is passed to the AI model
- Automated Code Reviews: The /review command does an automated review, focusing on potential bugs, security issues, and best practice violations
- Code Improvement Suggestions: The /improve command gives suggestions for optimizing your code
- Generate PR Descriptions: The /describe command creates PR descriptions including title, summary, step-by-step walkthrough and labels
The extension is free to use on public repositories. For private repos, we have PR-Agent Pro (14-day trial, then $19/month) with additional features like documentation suggestions, component analysis, and more.
I'd love to hear your thoughts and feedback. What features would make this most useful for your workflow? Any concerns or suggestions for improvement?
Chrome extension: https://chromewebstore.google.com/detail/ephlnjeghhogofkifjl...
Documentation: https://pr-agent-docs.codium.ai/chrome-extension/
PR Agent repo: https://github.com/Codium-ai/pr-agent
For the PR agent, i have a question, can I use it to review PR's that I've authored myself? I've recently contributed to the Bazel project and, man, these maintainers are busy so the idea is to use an agent to do a pre-review that covers the usual stuff only leaving the critical parts.
The other thing is, is your company entertaining the idea to parse issues into prompts and then implement them as to derive a PR? Looking at those rather popular repos, there are always a number of issues in the backlog that are a straight forward so I wonder why not going one step further?