Token burn
AI Agent Token Burn Rate Calculator
Token burn rate is the speed at which an agent consumes input, output, cached input, and tool-result context. Coding agents can burn tokens faster than chat because every turn may include files, diffs, test logs, and tool output.
Watch these signals
- Input tokens per agent turn.
- Output tokens per completed patch.
- Repeated context that should be cached.
- Tasks per day and active days per month.
Use the main calculator to convert token burn into estimated daily and monthly cost.