Current plan
$8,747/month- Per run
- $3
- Human review
- 96%
- Retry burn
- $1,141
- Idle burn
- $69
$69.31 per month is spent on scheduled runs with no work. Monitor it before volume grows.
Free AI agent cost calculator
AI agent cost calculator
Estimate tokens, tools, retries, and human review together.
Start the toolEstimated: 5 minutes
02
Your cost drivers
Per run, per month, and idle burn
The estimate separates model tokens, paid tool calls, and productive human review so a cheap model cannot hide an expensive operating process.
Baseline monthly estimate
$8,747
Human review is your largest cost driver.This is your baseline. Add a comparison to see how another setup changes the cost.| Period | Current plan |
|---|---|
| Model | $189 |
| Tools | $158 |
| Review | $8,401 |
What this means
Human review · not model tokens · is the biggest modeled cost. Test clearer acceptance criteria, narrower tasks, and better failure handling before switching to a cheaper model.
Show the agent-cost formula and assumptions
Monthly model cost = attempted runs × ((input tokens ÷ 1M × input price) + (output tokens ÷ 1M × output price)). Attempted runs include the selected retry share. Tool cost = attempted runs × tool calls × unit cost. Human review is charged only to productive attempts. Idle burn includes model and tool costs for the share of attempts with no work. The estimate excludes hosting, orchestration, storage, taxes, discounts, caching, and special processing tiers.
Export your current scenarios.
Unlock a Sheets-ready CSV with editable pricing, workload assumptions, cost breakdowns, idle burn, and the safe operating recipe. Your inputs stay in this browser.
- Your cost scenarios
- Largest operating cost drivers
- Editable CSV comparison
Preview the result or example included in this kit
Current plan is the baseline at $8,747.28 per month and $2.50 per scheduled run. Human review is the largest cost driver in this baseline. Add a comparison to test potential savings. Pricing reviewed 2026-08-29.
Maintained pricing
OpenAI GPT-5.6 standard API rates
Reviewed August 29, 2026 · Owner: Nerd Out product operations · Official model pricing ↗
| Tier | Model | Input / 1M | Output / 1M |
|---|---|---|---|
| Efficient | GPT-5.6 Luna | $0.20 | $1 |
| Balanced | GPT-5.6 Terra | $2 | $12 |
| Flagship | GPT-5.6 Sol | $4 | $20 |
Standard API processing under 270K input tokens. Excludes caching, batch, fast mode, regional processing, and provider discounts.
How this tool works
Workload and review assumptions in. Cost per run, monthly cost, the largest cost driver, and an idle-burn warning out.
- Enter the shared token and labor assumptions.
- Compare three model, volume, review, and idle-run scenarios.
- Get the editable scenario export and cost-control recipe.
The useful distinction
Model price is only one line in the agent bill.
Token cost can be the smallest part of an operating workflow. Paid searches and data calls repeat on every scheduled run, while even a few minutes of human review can dominate the monthly total.
Compare configurations against the same acceptance tests. A cheaper model is not cheaper if it creates more retries, review, or failed work. Measure idle runs too: event triggers and inexpensive pre-checks can prevent a schedule from paying to discover there is nothing to do.
Questions owners ask
AI agent cost FAQ
How much does an AI agent cost per month?
Monthly cost depends on run volume, input and output tokens, paid tool calls, retries, idle runs, and human review. This calculator models the first five directly; include retries in run volume and add hosting or orchestration separately.
How is AI token cost calculated?
Multiply input tokens by the provider's input rate and output tokens by its output rate, with each rate expressed per one million tokens. Add tool charges and labor to estimate the complete workflow rather than model usage alone.
Why can human review cost more than the model?
A few minutes of skilled review at every run compounds quickly. Better acceptance tests, narrower tasks, and clearer escalation rules may reduce review time, but required safety or quality checks should not be removed merely to lower the estimate.
What is idle burn for a scheduled AI agent?
Idle burn is model and tool spend on scheduled runs that find no useful work. Track the idle-run share and consider an event trigger, queue check, or cheaper pre-check before starting the full workflow.
How current are the model prices?
The visible pricing table shows its last-reviewed date and links to the official provider source. Prices can change, so verify the source before approving a budget and edit the rates in the exported scenario sheet when needed.