Replay & Time Travel
Debug and improve with confidence. Re-run any historical execution with updated prompts or models. Every workflow is reproducible by version hash. Measure improvements against real data. Debug production issues in your local environment.
# Replay a production run locally
$ agnt5 replay wf_abc123 --model=gpt-4o
✓ Replaying workflow wf_abc123
✓ Using model: gpt-4o
✓ Completed in 2.4s
# Compare results
$ agnt5 compare wf_abc123 wf_abc124
Diff: Output quality improved by 23%Reproducible by Hash
Every workflow execution has a unique version hash. Replay any historical run with the exact same code, inputs, and environment. Perfect for debugging production issues.
Test Prompt Changes
Iterate on prompts with real production data. Replay historical runs with updated prompts to measure improvements. No need to wait for new data.
Model Comparisons
Run the same workflow with different models side-by-side. Compare GPT-4 vs Claude vs Gemini on your actual production workload. Make data-driven decisions.
Local Debugging
Debug production issues locally. Replay any execution in your development environment. Step through with a debugger. Fix issues faster.