Skip to main content

Budget Alerts

AI SpendOps lets you set spend budgets and receive email alerts when usage approaches or exceeds them. Budgets are managed from the Budgets page in the dashboard and are available on all plans.

Budget scopes​

A budget can be scoped to exactly the slice of spend someone is accountable for:

  • Organisation — total spend across all API keys and providers
  • API key — spend for a specific key (e.g. production vs staging)
  • Provider — spend for one provider across all keys (e.g. Anthropic only)
  • Dimension — spend for one dimension value on an API key (e.g. team=data-science or app=chatbot), using the same dimensions that drive your cost reporting

Period types​

  • Calendar month — resets on the 1st of each month
  • Calendar year — resets on 1 January
  • Fiscal year — resets at your financial year start, as configured on your organisation

Thresholds and notifications​

Each budget has a critical threshold (typically 100%) and up to five warning thresholds below it, so you can shape the escalation: for example a heads-up at 50%, a warning at 80%, and the critical alert at 100%.

  • Alerts are delivered by email to the recipients configured on the budget.
  • Each threshold fires once per period and resets when the period resets, so a budget hovering around a threshold does not repeat-fire.
  • If spend crosses several thresholds between evaluations, only the highest level sends an email; the lower levels are still recorded in the budget's alert history.

Spend is evaluated against budgets every few hours.

Gauges, burn rate and projection​

  • The Budgets page shows each budget as a gauge with its current position in the period.
  • The main dashboard includes a This month's budgets panel.
  • Each budget's detail page charts cumulative daily spend and projects it forward at the current burn rate, telling you whether you are on pace to stay inside the budget or when you are likely to cross it.

Alerts inform, they never block​

Budget alerts are informational by design: they notify the people you choose, and traffic flows regardless. A budget crossing a threshold will never cause a request to fail. To enforce hard limits on what your keys can call, use provider and model restrictions, which are enforced at the edge on every request.