Hey HN, I built Vet. It's an Open Source code review tool, designed to work well with coding agents.

Coding agents frequently fail in subtle ways. They claim tests passed without running them, swap in fake data when they are unable to access resources, and silently half-implement features rather than admitting they are stuck.

Vet runs in the loop with your agent, reviewing conversation history alongside code changes to verify that what the agent did matches what you asked for. It catches logic errors, unhandled edge cases, and deviations from your stated goals. It reviews full PRs too.

No accounts, new subscriptions, or telemetry. Just bring your existing API keys and try it out!

Happy to answer questions!