Agent cost calculator

What will this agent cost to run?

Estimate what an AI agent costs to run: pick the model, say how often it runs and how many tokens a run uses, and see the cost per run, per day and per month.

How often the agent runs, on a schedule or on demand.

All the steps of a run together: an agent that calls tools sends its context again on each step.

What the model writes, across every step.

Claude prices cached input at a tenth of the input rate.

estimate · list prices
Per run
$0.0852
Per day
$1.70
Per month
$51.83

How it is worked out

List prices, applied the way your dashboard applies them.

Each model's published price per million input and output tokens, the same table AgentOS uses for its cost figures. Cached input is priced at a tenth of the input rate for Claude; for other models it is priced as fresh input, so for those models the estimate errs high rather than low.

Token counts are the part to measure. An agent that calls tools sends its whole context again on every step, so a run of ten steps uses many times the tokens of one reply. Once an agent runs on AgentOS, every run records its exact tokens and cost.

Cost and budgets in AgentOS

Questions

What people ask about agent costs.

Why does one run use so many input tokens?
An agent that calls tools sends its whole conversation to the model again on every step. A run of ten steps sends the system prompt, the tools and the history ten times.
Does AgentOS add a margin on tokens?
No. Agents run on your own model keys, and you pay the provider directly.
How do I cut the cost of an agent?
A smaller model for simple steps, fewer steps per run, a shorter prompt, and prompt caching where the provider offers it. The built-in Token Optimizer points out where tokens go.

Pilot

Try it on one real workflow.

We build the agent with you, with an approval gate on the risky step, and report what it did.

Templates

Start from an agent that already works.

45 templates for support, email, operations and reporting: install one, connect its tools, and gate the step that matters.

Measure it instead of guessing.

Run the agent once on AgentOS and read the real tokens and cost of every step.