AI spend

Inference spend for the org over a window (default the last 30 days): totals with the previous equal-length window for comparison, a time series by provider, and breakdowns by provider, by model and by agent — each agent with its shroud_config daily budget, today's spend against it and the requests Shroud blocked for budget. limits carries router-key spend caps, the credit balance and 1Claw's inspection fees for the window (a separate number from provider cost). Every request is priced at ingest from the price card (GET /v1/spend/ai/prices); a request whose provider has no card is counted as unpriced, not free. Human and platform callers only. Backs the dashboard's Spend tab; the same numbers reach an org's own collector as OTEL metrics ai.cost_usd, ai.tokens.input, ai.tokens.output.