Skip to main content
Spend limits protect you from unexpected usage. Set one budget for the organization, then give individual projects smaller caps where you need tighter control.
A $2,000 organization budget with a $500 project cap

Set the monthly budget

The usage indicator shows current month-to-date spend:
  • Green below the warning threshold
  • Amber at or above the warning threshold
  • Red at 100%
Warn at changes the indicator on this page; it does not send an email or notification.

What happens when you reach a limit

The call that reaches or crosses the limit completes. Opper blocks the next call with HTTP 402 Payment Required, so final spend can be slightly higher than the configured amount. Budget limits reset at the start of each calendar month in UTC. Adding credit does not clear a budget block because the limit applies to spend, not account balance. To resume calls sooner, raise or remove the budget; Opper re-evaluates the block as soon as you save.
A low account balance can also return HTTP 402. Adding credit clears a balance-related block, but not a monthly budget block.

Add project caps

Select Cap spend for a project to limit one project’s monthly spend. Project usage counts toward the organization budget—it is not an additional allowance. When you save, Opper checks that:
  • The organization has at most one enabled monthly budget.
  • Each project has at most one enabled project cap.
  • Every project cap is less than or equal to the organization budget.
These checks work in both directions. If an existing project cap is 500,forexample,youcannotlowertheorganizationbudgetbelow500, for example, you cannot lower the organization budget below 500 until you reduce the project cap.

What counts toward a limit

Spend limits cover every billable model call, including:
  • LLM and embedding calls through the v3 API and compatibility gateway
  • Image, speech, and video generation
  • Judge calls made by score checks
Use the organization budget as the hard ceiling for your account. Use project caps to contain workloads that may spike independently.