💰 Govern AI spend across teams
Stack — Scope: org / team / agent budgets · Caps: soft (alert) + hard (block) · Enforced: at run start · Visibility: Cost tab · Output: no runaway bills
Who it’s for — the person who has to explain the model bill, and wants a ceiling before it’s a surprise. What it does — sets spend budgets per team or agent with a soft cap that alerts and a hard cap that blocks. The limit is checked when a run starts, so a looping agent or a bad prompt can’t quietly burn the quarter’s budget.
Build it
-
Decide the scope. Budgets attach at the org, team, or agent level — give each team a monthly ceiling, or cap an experimental agent tightly.
-
Set a soft cap — an alert threshold. Crossing it notifies you but lets work continue.
-
Set a hard cap — the enforced ceiling. When spend would exceed it, the next run is blocked rather than billed.
-
Watch the Cost tab — spend per agent and team, so you see which agents actually drive the bill (a cross-vendor verify panel earns its cost on a contract reviewer, not a formatter).
-
Combine with BYOK — bring your own provider keys for the volume you’d rather meter directly, and let budgets bound the rest.
Make it yours
- Cap a customer-facing agent per tenant so one account can’t exhaust shared budget.
- Alert to Slack on the soft cap; page on the hard cap.
- Pair with analytics to attribute spend before you set the numbers.
Uses — Enterprise · Cost governance · Budgets