Governance policies

Why governance

Agents do real, metered work — which is exactly why an organization wants guardrails around them. Governance policies, at Organization settings → Governance (owners and admins, Enterprise plan), put two controls in your hands: monthly spend budgets and a model allowlist. Both are enforced by the platform before work starts, not reported after the bill arrives.

Spend budgets

Set a monthly USD budget per member, per project, or both. Budgets apply to attributed usage — agent runs, workspace sessions, and the other metered work each person or project generates — and reset every calendar month. Leave a field empty for no limit.

Enforcement is automatic: when a member or project reaches its budget, new agent sessions, chat runs, and deploys for it are refused with a clear error until the month rolls over or an admin raises the budget. Work already deployed keeps running — budgets stop new spending, they don't take apps down.

Budgets cap spending; they don't grant credits. The organization's usage still bills against plan and prepaid credits as usual — budgets decide who may consume them, and how much.

Model allowlist

By default, members can use every model in Spaceboy's curated catalog. The Model access panel restricts that: uncheck the models you don't want offered, save, and agents and project setup only present the allowed subset from then on. At least one model must remain allowed, and Allow all models clears the policy entirely.

Auditing

Every policy change — budgets updated, the allowlist changed — and every policy-triggered refusal is written to the audit log under the Policy category. When someone asks why their run didn't start, the answer is on the record.