Hey HN! Ilya and Nikita here. We're working on wispbit - the open source AI code review agent for teams - https://github.com/wispbit-ai/wispbit. We made wispbit because we previously struggled in big codebases and teams that grew quickly. Keeping codebase standards alive was frustrating, and investing in static internal tooling was too hard. We would sit in post-mortem meetings and expect engineers to remember every single booby trap. wispbit fixes this by enforcing your codebase rules, and raises a violation if a rule is broken. It also runs anywhere (Github actions, CLI, Claude code). You can come up with rules on your own, or use our free rules repository https://wispbit.com/rules. Some ways we and other engineers use wispbit: - Enforce database migration or schema patterns - consistent table names, columns, checking for redundant indexes. - Take a look at your most recent postmortem doc and turn it into a rule. - Enforce standards for commenting, test writing patterns, and component usage to keep your team accountable. - Take something you frequently check for and turn it into a rule. Why wispbit over other tools? We found that existing code review tools are too random and noisy - a level that is unacceptable in big codebases and teams. wispbit keeps it simple by reviewing only what you ask for. We're also working on a cloud version over at https://wispbit.com, currently in early access. It focuses on creating and maintaining rules for you based on codebase activity. In the long term, we want to fix tribal knowledge and reduce the burden on teams having to remember every single detail within complex stacks. If this resonates with you, or you built your own code review tool internally - give it a spin! |