Diagnose Your AI Agent
Configure the engine, load an execution trace, and quantify Schema Re-parse (A), Cascade (B) and State Growth (C) cost. The whole analysis runs in this browser tab.
Engine settings
Baseline factors · GPT-5 / 5.1. Open to configure the tokenizer, token, compute, electricity and carbon costs.
Execution trace
Parsed locallyDrop an agentic execution CSV, or tap to browse
One row per step. Step, payload-token, failed-flag and re-parse columns are auto-detected.
Steps
No steps yet. Import a CSV above, or add steps one at a time.
Results are indicative. Every figure depends on the engine settings you choose above and the trace data you load. Please confirm both are configured correctly before running the tool.
Add or import at least one step to run the diagnosis. Nothing you enter is uploaded to a backend.