Pay for the work. Not for seats.

It depends on the workflow. Book your discovery call: we look at the work with you and quote it before anything runs.

The time a run spends planning and doing the work, billed by the second. Time queued behind another run on the same computer, or waiting for your approval, an answer or a sign-in, isn't counted.

Per token, at a rate that depends on the AI model. A fast, light model costs much less than a frontier model, and you choose the model for each role in Settings. Fleet reads your apps directly and keeps most of each request cached, which keeps usage low. Each run's page shows what it used and what it cost.

Yes, on request. With your own key from Anthropic, OpenAI, Google or OpenRouter, your provider bills you for the AI directly and Fleet AI charges only the runtime. Book a call and we'll set it up with you.

When you'd rather have one agreed price than a usage bill, usually for recurring work with a steady volume, like a monthly reconciliation or a weekly report run. Tell us the job and we'll quote it.

Yes. Give a run or a workflow a spending limit: when it's reached, the run pauses and waits for you, and nothing more is spent while it waits. Each AI app you connect over MCP can also have a monthly cap, after which it can't start new runs.

Runs draw from your account's prepaid credits. When the balance runs out, new runs don't start until credits are added.

It's billed for what it used before it stopped. A run that can't continue, for example because the computer's session dropped, stops right away instead of retrying, and its page says why.