Skip to content

AI Spend Caps: overview

Someone on your team uses an AI feature all afternoon, and nobody sees what it cost until the figures come in. Here you set one monthly budget that covers every site and a daily limit for each person. Once either one is used up, Capell stops further AI requests. AI Spend Caps works on top of AI Orchestrator, which runs the requests.

When spending reaches the share of the budget you choose, the person you name gets one email for that month. An editor who hits a limit sees a short message saying which one it was, so they know whether to wait or ask an administrator to raise it.

An AI usage page in the admin shows where the money went, split by AI task, by person and by site. The figures are estimates, worked out from how much text each request used and a price list your developer keeps up to date. They won’t match an invoice to the penny. If you’d rather watch than block, turn the hard stop off and keep the warning and the usage page.

Spend limits for AI features. It adds a monthly site budget, a daily limit for each person, one warning email and an AI usage page.

It works on top of AI Orchestrator. Every AI request passes through a guardrail that adds up the estimated cost so far and refuses the request once a limit is reached and the hard stop is on.

Go to AI → AI usage to see estimated spend for the month by model, task, person and site. Limits are set in the AI Spend Caps settings.

Monthly budget and daily limit per person are in US dollars; a value of zero means no limit. The warning percentage and the warning email address control the one email sent each month. Turn off the hard stop to keep the warning and the page without blocking requests.

  • Costs are estimates. They come from the tokens each request used and the price list in the package configuration, so they can differ from your provider’s invoice.
  • A refused request leaves content unchanged and tells the person which limit applied.

For developers: see the README.