Skip to main content
Wardin’s budget model lets managers cap governed, gateway-routed AI spend at two levels — per key and per user — enforced atomically before any token is spent.

Budget hierarchy

Both levels are checked independently. A request is blocked if either limit is exceeded. There is no separate tenant/org-wide budget tier.

Step 1: Set up team keys

Create one key per developer or team role:
See the full endpoint reference: POST /v1/keys.

Step 2: Configure alerts

Set a webhook alert so you’re notified before budgets run out: Console → Webhooks → Add endpoint
See the full endpoint reference: POST /v1/webhook-endpoints.

Step 3: Distribute keys

Send each developer their wardin_sk_ key. They set it as ANTHROPIC_AUTH_TOKEN (and ANTHROPIC_BASE_URL=https://gw.wardin.ai) in their environment. No other changes to their workflow.

Budget increase requests

Developers can request a budget increase from MY LANE in the dashboard. Requests land in LEDGER, where you can approve, deny, or set an expiry date for temporary increases.
  • Approved with expiry: The limit is elevated immediately and automatically reverts at the expiry date.
  • Approved permanent: The limit is raised and stays until you change it.
  • Revoked: You can revoke an approved increase at any time before expiry.
The full request history is retained as an audit trail.

Viewing spend

Overview shows org-wide spend and allocation; LEDGER breaks it down by team and developer. Each developer sees their own sessions with per-session cost under MY LANE. To pull spend data programmatically:
See the full endpoint reference: GET /v1/usage — query params and response schema. Start conservative — developers can request increases self-serve, and approvals apply instantly.