Routing and budgets

# Routing and budgets

Routing and budgets are workspace-level controls. They govern how requests are dispatched to models and how much a workspace can spend — enforced by the platform, not by code you have to write and maintain.

## Routing policy

Routing policy is enforced at the API layer, not in your application code.

You declare the policy — which models are eligible, cost ceilings, latency tolerance, residency constraints — against the workspace, and every request inherits it. Your application keeps making ordinary OpenAI-compatible calls.

## Budgets

Set a budget per workspace. When it is reached, further requests are held until you raise it -- no surprise invoices.

Budgets are a spend cap, not an alert. Holding requests at the API layer means a runaway job or a misbehaving agent cannot spend past the ceiling you set.

## Auditable ledger

Every routing decision is recorded, so usage and residency posture can be verified after the fact.

The ledger is the source of truth for what ran where and at what rate — useful for cost attribution and for demonstrating residency posture to reviewers.