Usage, credits & quotas
What your tenant is consuming against what it is entitled to: LLM credits, provisioned disk, and Agent Cloud nodes.
Three allowances bound a tenant: credits for model tokens on the platform-default model, a disk quota for provisioned storage, and a node quota for managed Agent Cloud capacity. All three are live figures under Manage → Usage, and the credits meter is mirrored in the sidebar.
The Usage page
Four headline tiles — credits available, credits used, nodes deployed, disk used — over meters showing each figure against its limit. Bars turn amber past 80% and red at the limit, so an allowance about to bite is visible before it does.
Credits
A credit is a token bucket: 1 credit = 1,000,000 tokens of combined input and output on the platform-default model. New tenants are seeded with a starting allowance. Credits are consumed by platform features that call an LLM for you — generating a policy or a skill, the dashboard composer — and by agents running on the platform-default model.
At zero credits, features that need the platform LLM stop and report out_of_credits rather than failing obscurely.
Disk & node quotas
| Quota | Counts | Enforced when |
|---|---|---|
| Disk (Gi) | Provisioned file systems and app storage across the tenant. | Creating or growing a volume. |
| Nodes | Nodes running managed Agent Cloud deployments. | Creating or resizing an Agent Cloud deployment. |
A request that would cross a quota is refused with the limit that was hit — and Agent Cloud sizing reports every violation at once, not just the first, so you can correct the whole form in one pass.
Raising an allowance
Entitlements are set per tenant and read by anyone with usage read access. Changing them needs usage write access. Values are absolute, not deltas — adding 100 credits means writing the current total plus 100.
1GET /api/v1/dashboard/limits2PUT /api/v1/dashboard/limits { "credits": 250, "disk_limit_gi": 500, "node_limit": 10 }Per-agent spend
Credits and quotas are tenant-wide. For attribution — which agent, which person, which day, and what it cost in dollars — route traffic through an LLM gateway. Its usage view slices spend, requests, and tokens by agent and by user, and its budget rules cap them before the tenant allowance ever comes into play.