Replaying Executions
Step through recorded agent executions like a video with the Tape Player.
The Tape Player
Every recorded trace can be replayed in the Retrace Tape Player. Navigate through each step of your agent's execution, inspect inputs and outputs, and understand exactly what happened.
Navigation
The player has three panels:
| Panel | Position | Purpose |
|---|---|---|
| Span List | Left | Chronological list of all steps |
| Content Viewer | Center | Input, output, and metadata for selected span |
| Timeline | Bottom | Proportional duration bar with playback controls |
Playback Controls
| Control | Action |
|---|---|
| Play/Pause | Auto-advance through spans |
| Skip Forward | Jump to next span |
| Skip Back | Jump to previous span |
| Speed (0.5x-4x) | Control auto-advance speed |
| Timeline click | Jump to any span directly |
Span Types
Spans are color-coded by type for quick visual scanning:
| Color | Type | Examples |
|---|---|---|
| Purple | LLM Call | retrace.ai.generate |
| Amber | Tool Call | web_search, write_file, api_request |
| Red | Error | Exceptions, timeouts, rate limits |
| Green | Fork Point | Where a fork branches from |
| Violet | Reasoning | Chain-of-thought, planning steps |
Sharing Tapes
Publish any trace as a shareable tape:
- Open a trace in the dashboard
- Click "Publish as Tape"
- Set visibility: public (anyone), unlisted (link only), or private (you only)
- Share the URL - recipients can replay without logging in
[!TIP] Unlisted tapes are perfect for bug reports and code reviews. Only people with the link can view them.
Use Cases
- Bug reports: Share the exact execution that failed
- Code reviews: Show how an agent reasons through a problem
- Documentation: Demonstrate agent capabilities with real examples
- Debugging: Replay a user's exact execution to reproduce issues