Example inputs: 10,000 tasks/month at the default 15 turns/task and 70% success rate
Pricing last updated 2026-05-28 — verify with provider before committing to commitments or contracts. Agent cost per successful task: $0.3600 (15 turns × ~3600 in / 400 out per turn on anthropic/claude-sonnet-4-6). Per-attempt cost: $0.2520; success rate 70%; failure overhead adds 30% to effective cost. At 10,000 tasks/mo: $2520/mo total, 7000 successful.
Example inputs — your own numbers will differ. This is a computed planning estimate, not a quote.
Enter tasks per month, steps per task, tool calls, token sizes per turn, and success rate; Model Ruler multiplies loop depth and fan-out and divides by success rate to report cost per successful task and monthly cost. Reducing loop depth is usually the first lever once production traces are available.
Models cost per successful agent task by multiplying turn count, tool-call fan-out, context growth, and failure overhead.
Formula: cost_per_successful_task = cost_per_attempt / success_rate; monthly_cost = tasks_per_month * cost_per_attempt
Audit coverage: Agent-loop fan-out · Context growth · Observability overhead
What does this calculator compute?
It multiplies turns, tool-call fan-out, and per-turn tokens, then divides by success rate to report cost per successful agent task and monthly cost.
What should I check before committing to a workload shape?
Each loop iteration re-sends accumulated context, so cost compounds with the number of steps and the growing context per step. Check the expected iteration count and whether context is trimmed between steps — an unbounded loop on a large model is where agent cost runs away.
Is this a quote or a benchmark?
Neither — it is a computed estimate, not a vendor quote. The worked example above shows the numbers this calculator produces for a representative agent loop cost workload; your own inputs move the result.
Estimates use published per-token API pricing for the model/provider path shown by the calculator, reviewed on the source date shown. Provider rates change frequently; verify current pricing on the provider's own page before committing to traffic volume, eval cadence, or agent-loop workload shape. View methodology.