About the Agent Cost Simulator
A single-turn chat request and a multi-step agent loop have very different cost profiles — every tool call, reasoning step, and retry consumes tokens, and retries compound faster than people expect. This tool projects per-run and monthly cost from a step count, average tokens per step, and a retry rate.
Key Features
- Models retry overhead as effective extra steps, not just a flat multiplier on the happy path.
- Per-run and monthly cost at your actual traffic volume.
- Same model pricing table as the Token Counter.