agent-replay provides time-travel debugging and side-by-side diffs to pinpoint exactly where AI agents hallucinate or fail. It replaces manual log diving with a local-first toolkit to replay, fork, and automatically evaluate agent traces for faster iteration. It's a work-in-progress. I'd love any feedback. Thank you.
Show HN: Time-travel debugging and side-by-side diffs for AI agents | Heykuki News