Configure an agent run as you would build it with the GDK: model choice, prompt and completion tokens, tool calls, memory steps, and loop iterations. Every change recalculates estimated token spend, cost per run, and projected monthly usage instantly. Estimates are illustrative, not real billing.
Run Configuration
Token accounting caveat: real agent runs vary. Providers count system prompts, tool schemas, and cached context differently; retries and truncation change totals. Use this simulator for order-of-magnitude planning, then verify against provider billing dashboards.
Live Estimate
0
Tokens per run
$0
Cost per run
$0
Cost per day
$0
Projected 30-day cost
Input vs output tokens per loop iteration
Teal = input (prompt, memory, tool responses). Indigo = output (completion tokens). Context grows each iteration because prior turns are re-sent.