Spend Control

AI spend control teams can actually buy

Customers need more than model access: they need to know what every API key, customer, feature, and model costs, and cap risk before spend runs away.

Workspace

Monthly budget

API key

Budget / daily cap

Request

Max cost

Attribution

customer_id / feature

From API-key budgets to workspace budgets

Existing API-key monthly budgets stay in place; Spend Control adds workspace budgets, per-request max cost, and model allow/block controls.

  • Budget hits return customer-safe 402/429 responses with remaining-budget headers.
  • Model allowlist/blocklist is inherited from workspace to request execution.
  • Daily caps and per-request max cost limits prevent abnormal traffic from expanding losses.

Attribution goes into the ledger by default

Customers can pass customer_id and feature through headers or payload metadata for usage, billing, and anomaly review.

  • Supports X-BatchIn-Customer-Id and X-BatchIn-Feature.
  • Attribution fields are stored on UsageLog and billing metadata.
  • Anomaly alerts use budget, growth-rate, and model-switch signals.