Open-source AI code review (AGPLv3, BYOK)
BYOK model avoids vendor markup, but GitHub Copilot already dominates this space.
Reviewer Agent you can self-host
Uses your own Codex or Claude sub to avoid per-seat SaaS pricing.
Indie developers and small teams with existing AI subscriptions
Reviewbot · CodeRabbit · GitHub Copilot
I wasn't happy with the code review tools we use at work and on my side projects. Noisy, reviews in surface, expensive (overkill for sides). So since Gemini was getting sunsetted, I explored having my own.
As I am a cheap rat, this weekend I built Themis, that works with my Codex/Claude subscriptions (I understood "claude -p" is ok again in subscriptions).
I wanted it open-source so it can help other indie devs or small teams, self-hostable (or local), drawing from subs and mostly, configurable so it works well with your own repos.
No big ambitions nor 1 billion startup here, just fun (so please don't roast ahah). If it is useful to even a few people, that's already great.
Happy to have feedback (especially on distributing the app!)
Ps: 1st HN contrib, hope it's the right format
BYOK model avoids vendor markup, but GitHub Copilot already dominates this space.
Self-hosted AI review with CVE tracking beats Greptile on speed and privacy.
Model-agnostic code reviews with zero LLM markup beats Claude Code Review on cost.
Claude in GitHub Actions with inline comments, but Codiumai and others already own this niche.
GitHub PRs to performance reviews with local Ollama support for privacy.
Cross-model decorrelation catches blind spots same-model reviewers miss.