Time-travel debugging for AI agents.
We are building local-first tools to record, inspect, replay, fork, and learn from AI agent executions.
Agent Replay is a debugging layer for AI agents that helps developers:
- record structured agent execution traces
- inspect LLM calls, tool calls, retrieval events, and memory events
- replay failed runs locally
- fork execution from any step
- compare alternative prompts, models, tools, and retriever configurations
- turn successful and failed trajectories into reusable experience memory
Trace → Inspect → Replay → Fork → Learn
