Hey everyone, I'm David, the developer behind Kakarot CI. I built this because I kept watching the same pattern play out. Developers were already using AI to generate unit tests, copying functions into ChatGPT or asking Cursor to write them. The workflow was slow, the prompts were inconsistent, and the output was hit-or-miss. Half the time the tests didn't even run. Import paths were wrong, mocking was off, types didn't match. Developers were spending more time fixing AI-generated tests than they would have spent writing them from scratch. I figured if developers are already going to use AI for testing, why not make a tool that does it properly, with structured prompts, real AST analysis of the code, actual test execution, and automatic fix loops when things break. No more copy-paste workflows. What surprised me during development was how useful the local mode became. I originally designed Kakarot CI strictly as a CI tool. It would run in GitHub Actions, generate tests for PRs, and commit them automatically. But while building it, I kept running the CLI commands locally against my own code changes and found it genuinely useful for day-to-day development. So I expanded the CLI to be a first-class experience alongside the CI workflow. This is a solo project, built as a side project over several weeks. It's currently in beta and free to use, you bring your own LLM API key. I'd appreciate any feedback, especially from developers who are already using AI for testing and have felt the friction I'm describing.

Kakarot CI

Stop writing tests. Start shipping code.

Stop writing tests. Start shipping code.