# AI agent cost estimator: run it free, or from your agent

Cost per task and per month for an agent that loops, counting the history it re-sends on every step.

## Prompt

Use looot to price my AI agent: [model], [fixed prompt tokens] prompt tokens, [tool result tokens] tool tokens and [output tokens] output tokens per step, [steps] steps per task, [tasks] tasks a day. 1. Set up https://looot.ai/skill.md. 2. Run the OpenRouter model list endpoint for today's prices. 3. Sum the input of every step, history included. 4. Return cost per task and per month for my model and three cheaper ones that fit the context.

## What it needs

A model, prompt and tool result sizes, steps per task and tasks per day

## What comes back

- cost per task, day and month
- input and output tokens per task
- input size of the last step
- cheaper models whose context fits

## Endpoints and providers

None. The tool on the web page runs in the browser and calls no endpoint.

## What it costs

Nothing. The tool on the web page runs in the visitor's browser: no account, no API call and no limit per hour. The prompt above runs the same job from your own looot account with current data, billed per call.

## Budget an agent

Know the monthly bill before you put an agent in front of users.

## Cut steps

See how much one fewer tool call per task saves at your volume.

## Compare models

Switch to a cheaper model whose context still fits the last step.

## Catch context overflow

Get a warning when the last step outgrows the model's context window.

## How much does an AI agent cost to run?

Add up every model call in a task. Each step sends the fixed prompt plus all earlier tool results and outputs, so cost grows faster than the step count. The estimator does that sum for you.

## Why does each step cost more than the last?

The agent keeps its history, so step ten sends everything from steps one to nine again. The page shows how much of the input is re-sent history.

## Does it include prompt caching?

No. The model list has no cached input prices, so the estimate assumes no cache. With caching on, a real agent costs less than shown.

## Where do the prices come from?

From the OpenRouter model list, read through looot on the date shown under the estimator. Check that date before you quote a number.

## What counts as a step?

One model call. An agent that searches, reads two pages and answers makes four calls, so four steps.

## How do I make my agent cheaper?

Fewer steps, shorter tool results and trimming old turns from the history all cut input tokens. A cheaper model helps when its quality holds on your tasks.

## Is the estimator free?

Yes. It runs in your browser with no account and no limit.

## Related tools

- /free-tools/llm-token-cost-calculator
- /free-tools/token-counter
- /free-tools/ai-model-comparison

## Last reviewed

2026-10-14.