How does it decide what to test in a PR? And how does this code review tool connect with services that require configs? E.g. My checkout uses stripe, and for QA, we have a sandbox account connected.
It uses code diffs + descriptions + issues as the base test plan (acceptance criteria + regression). You can then interact with it to shape it more towards what you want.
Static code analysis is great but it misses a lot. Agents that read code make guesses about how code would execute, Ito makes a guess and then verifies it with runtime evidence.
Our first POC was slow, expensive, and bad quality. It takes a lot of effort to build the infrastructure, memory layers, QA code gen fixtures and more to get it working reasonably well.