Agent-generated code can arrive faster than a developer can build a mental model of it. The real risk is not just a noisy diff: it is losing track of why a change was made and how its design maps to the implementation. This project connects diagrams, code, and agent decision traces so review can focus on understanding a change rather than merely scanning lines.
What Sets It Apart
- Diagrams link to implementation: Sequence diagrams, entity-relationship diagrams, and trace excerpts can point to the code behind them. This makes architectural questions easier to check against actual behavior.
- Diffs are organized around meaning: An AST-aware viewer can summarize large added functions and fold or hide less relevant changes, such as tests and documentation. Reviewers can spend more time on consequential behavior.
- Agent decisions become reviewable: Requirements and autonomous choices can be linked to the work produced, helping teams spot when an implementation diverges from the intended design.
Who It’s For—and the Trade-offs
A good fit for developers using coding agents who need to review architectural changes, prototypes, or large diffs without losing sight of the reasoning behind them. Look elsewhere if you need to edit files in the review app, routinely review changes spanning multiple repositories, or depend on live updates to reviews already shared with teammates; those workflows are currently limited.
