ABOUT
Codebase intelligence you can check.
repowise indexes a repository once and serves that index to your agents, your editor, your pull requests and your team. Everything we claim about it is published with the method that produced it.
HOW WE WORK
Three things you can verify
An engineering leader evaluating tools has been told a lot of unfalsifiable things. None of the below is one.
Every published figure carries its sample size, its corpus and its caveat. Code health scores 0.74 ROC AUC at predicting which files get bug-fixed over the next six months, across 21 repositories and 2,826 files, with a 95% confidence interval we print next to it.
Read the benchmarksIndexing is 22x slower than CodeGraph on a like-for-like run, and the benchmark page says so under a heading that says so. If a call graph is all you need, CodeGraph builds one faster and we will tell you that before you pay us anything.
See the row we loserepowise 0.54.0 is open source and self-hostable. The 25 defect markers behind the health score are deterministic and run locally with no LLM and no cloud, so you can read the code that produced your score.
Read the sourceWHO BUILDS IT
The people behind repowise
Small team, public work. The commit history is the resume.
Co-founder, and author of the defect-prediction validation study behind the code-health score.
Read the profileCo-founder. Writes about learning a codebase you did not write, and what coding agents should know before they change one.
Read the profilerepowise is developed in the open. Issues, discussions and pull requests all live on GitHub.
Open the repository