Costs & Usage

Every AI call your workflows make runs on your own connected provider keys, so the money spent is your real spend, not a 60 Minute Creator fee. This page shows you where to see what you've spent, broken down by workflow and model.

The Dashboard

Your spend lives on the Dashboard, the page you land on after logging in. There is no separate "Costs" tab in Settings, everything is here.

The Dashboard greets you by time of day, then shows the title "Dashboard" with the subtitle "See how much your AI usage costs, by model and workflow."

Choosing a time range

A range selector lets you switch the whole Dashboard between four windows:

Today: the current day
7D: last 7 days (the default)
Month: the current calendar month
All: all time, month by month from your first spend

Every card on the page updates when you change the range.

What you'll see

The Dashboard shows four cards:

Headline card: your total spend for the selected range, your average spend per workflow, and (once you have runs) your single most-used workflow's own spend as a clickable chip. The ? next to the total explains why it can differ from what Atlas Cloud shows.
Spend over time: a full-width chart of your spend day by day (or month by month, for "All time").
By workflow: a ranked list of your workflows by spend. Click a row to jump straight into that workflow. Spend that didn't come from a workflow, like the Creative Studio, shows as "Outside workflows". A workflow that no longer exists shows as "Deleted workflow".
Top models: a ranked list of the AI models used most, each showing the model's own icon plus a small icon for every provider that billed calls for that model.

If a card has no spend to show for the selected range, it explains that plainly instead of showing an empty chart, for example "No spend in this range." If the cost data can't load, you'll see "Couldn't load cost data" with a suggestion to try switching the range or refreshing.

How your spend is calculated

Every amount you see is what Atlas Cloud actually charged you. The app never shows an estimate.

Images, video, voice and lipsync: every job has its own price at Atlas Cloud, and that's the charge you see. Most jobs have it the moment they finish. When Atlas Cloud posts it a little later, the app reads the job again until the price is there. A job Atlas Cloud blocks for its content can still cost a little, and the app records that charge too.
Text (LLM, Agent, Screenwriter): Atlas Cloud bills text by tokens, at the token prices on each model's page. Each reply is priced from its tokens the moment it finishes, which is exactly what Atlas Cloud charges for it. After Atlas Cloud closes the day (midnight UTC), the app checks the day against your Atlas Cloud billing. A reply that comes back without its token count gets its share of that day's bill instead. Only charges made with the key you saved in the app count.

Amounts show two decimals, cut off rather than rounded up, so $0.036 shows as $0.03. An amount under one cent shows three decimals, for example $0.009.

Why your total can differ from Atlas Cloud

If the total doesn't match your Atlas Cloud billing page, it's one of these:

An image or video price can take a few minutes to arrive.
Days here run on UTC. Atlas Cloud shows your local time.
Only spend with the key saved in this app counts.
Amounts are cut off, not rounded up: $0.036 shows as $0.03.

Capturing cost never blocks or fails a workflow run. If cost capture has a problem, your run still completes normally.

Per-run cost

Every run in a workflow's run history shows its own total cost, plus a small chip for each provider that was actually billed in that run (for example, an Atlas Cloud chip). A run with no cost shows no chips.

If a run includes a charge Atlas Cloud hasn't reported yet, the run shows "Atlas Cloud hasn't reported this charge yet" instead of an amount. An image or video price usually arrives within minutes.

There is currently no cost breakdown by individual node within a run. Cost is shown per run (with the provider split above) and account-wide on the Dashboard, but not broken down node by node.

Using an AI assistant to check your costs

If you've connected an AI assistant through MCP (see Settings, MCP Server), it can read the same cost data on your behalf: totals, trends, and breakdowns by workflow, model, or provider. Like the Dashboard, this does not include a per-node breakdown either.