Skip to main content

Budgets

Cap how much a team can spend on inference by setting a budget on its LLM gateway policy. You manage budgets from the policy's Budgets tab.

Set a budget​

Budgets are per policy: you set one policy-wide budget rather than a limit for each user, which makes them a natural fit for managing a team's spend. From the Budgets tab you can set a budget, then remove, update, or reset it as your needs change. A reset applies to the current period.

What happens when a budget is exceeded​

Once spend passes the budget, the policy stops accepting new inference. In MintMCP you see the policy as over budget:

An LLM gateway policy's Budgets tab showing estimated utilization over its monthly capacity, flagged as Exceeded

Agents running under the policy get a 429 response on rejected calls:

Claude Code showing a 429 error: the monthly budget has been exhausted

Auditing​

Every budget change is recorded in the audit logs, so you have a trail of who set, updated, or reset a budget and when.

Notes and limitations​

  • Budgets are estimates, not provider bills. MintMCP computes spend as calls happen, so there can be small differences from your provider's final invoice.
  • Limits trigger after they are passed. A budget stops traffic once spend has crossed the limit, so the call that crosses it still goes through.
  • Standard model pricing. Budgets use standard published model pricing. They do not account for provider-specific negotiated token rates, so a policy with a discounted contract will show higher spend than you are actually billed.