AI Agent Task Cost Calculator
Token calculators price a single prompt. An autonomous agent chains many tool calls to finish one task, multiplying the real cost. Enter your agent's step count and token usage to see the true per-task, daily, and monthly cost. Nothing uploaded.
Cost
Multiplier Effect
Learn more: what multi-step agent costs actually add up to
Why the multiplier always matches your step count
This calculator applies the same average input and output tokens to every step you enter, so the multiplier is never more than a restatement of the step count. Take the default numbers: 8 steps at 2,000 input and 500 output tokens each, priced at $3 per million input tokens and $15 per million output tokens. Each step costs $0.006 in input plus $0.0075 in output, for $0.0135 a step.
Eight of those steps come to $0.108 per task - exactly 8 times a single step's $0.0135, which is what the multiplier shows. At 100 completed tasks a day (3,000 a month), that scenario works out to $10.80 a day and $324 a month.
The same ratio holds at any step count. A 20-step task on those same per-step numbers costs $0.27, exactly 20 times a single step - not "well over 20 times," since the model treats every step as identical.
Where a real agent's bill can outrun this multiplier
Real agentic loops don't always keep each step's token count flat. Simon Willison's writeup on how coding agents work explains that most tool-calling loops replay the entire conversation so far on every turn, so "as a conversation gets longer, each prompt becomes more expensive since the number of input tokens grows every time."
This calculator doesn't model that growth for you. If your agent's context expands step by step, enter the average input tokens across the whole task rather than the count from an early step, or your real bill will run higher than the number shown here.
FAQ
Why is my agent so much more expensive than a single prompt?
You pay for every step, and each step's input and output tokens cost the same as a single prompt's would. An 8-step task costs exactly 8 times a one-step task with the same per-step token counts - not more, not less, because the calculator assumes every step looks the same.
Does the per-step cost grow as an agent runs longer?
Only if you tell the calculator it does. Many real agents resend a growing conversation history on every tool call, so their actual per-step tokens climb as the task goes on. Enter the average across the whole task, not the smaller number from an early step, to capture that.
How can I reduce my agent's monthly cost?
Cut steps, cut tokens per step, or both - the cost scales directly with each. Trimming or summarizing older conversation turns keeps per-step input tokens from growing, using a cheaper model for routine steps lowers the price per token, and caching repeated context (where your provider supports it) reduces what you pay for tokens you've already sent once.