Show HN: Build your own AI code reviewer
We built this because we worked in big and complex codebases where we kept hitting booby traps - often the same ones. People forgot things, or quit altogether, amplifying the problem.
We looked for other ways to fix this, but the solution is usually a combination of: - Writing a linter rule - too difficult and time consuming. - Writing docs and having frequent meetings on alignment - basically a full time job. - Using plug and play code reviewers - too generic and chatty.
So we built wispbit. The premise is simple: 1. Connect your repo and wispbit will generate some codebase rules based on repo activity, or add your own 2. Run the reviewer in the CLI to get early feedback before opening a merge/pull request 3. Reviewer also runs in CI to enforce the rules for everyone
Our thesis is that putting your "tribal knowledge" into a set of rules reduces downtime/unplanned work, ramps up engineers faster, and aligns your team on what kind of code is acceptable.
wispbit is not only a companion for humans, but for LLMs generating code. With the rise of “vibe coding”, many engineers (ourselves included) are starting to catch themselves merging in code without reading it.
Right now, wispbit will do some heavy lifting for you, but it’s still up to you and your team to write and maintain the rules. We imagine a future where we come up with rules before you even think of them - by digesting your Slack, Notion, and Linear/Jira activity, since that’s where most of the technical discussions occur.
We would love to hear thoughts from anyone that is involved in the code review process! We've made it free to use while this post is up.