Token Budget

Know what a task will cost before the agent starts.

Set context size, expected output and iterations. Edit the prices to match your model. Values are saved in this browser.

Example prices per million tokens. Replace them with your provider's current rates.

Scenarios

Lean, normal and worst case token budgets
CaseIterationsInput tokensOutput tokensCost

How it works

The model behind the numbers

Lean uses 60% of the context and 60% of the iterations, with no retries. Normal uses your inputs plus the retry allowance. Worst case uses 160% context, double the iterations and at least 50% retries.

Paper-collage of three towers of paper strips of different heights beside paper coins and an hourglass.

Cost =input tokens × input price + output tokens × output price, per iteration, multiplied by iterations and retries. These are planning estimates, not invoices.

Next: decide if the task is safe to delegate.