Retrace records AI agent executions, showing every LLM call, tool invocation, and error, with replay, fork, and sharing workflows. Developers can rerun from a branch point, compare divergence, and decide whether a change improved or regressed agent behavior.
Features
- Agent execution recording
- Run replay
- Fork runs
- LLM call tracing
- Tool invocation and error inspection
- Run sharing
Use Cases
- AI agent debugging
- Regression localization
- Prompt and toolchain iteration
- Agent observability
- Run reproduction and sharing
- Model nondeterminism analysis
FAQ
Retrace records AI agent executions, showing every LLM call, tool invocation, and error, with replay, fork, and sharing workflows. Developers can rerun from a branch point, compare divergence, and decide whether a change improved or regressed agent behavior. Core capabilities include: Agent execution recording, Run replay, Fork runs.
Common scenarios include: AI agent debugging, Regression localization, Prompt and toolchain iteration.
