How to control costs and set budgets
Per-run budgets, per-tenant quotas and rate limits, and dashboards that attribute every token.
Agentic compute is bounded before it runs and attributed after it finishes.
Per-run budgets
Every agent declares hard ceilings in its descriptor: a maximum duration, a maximum number of tool calls, and a token budget. A run that hits a ceiling stops with an explicit error — it never runs open-ended.
Per-tenant quotas and rate limits
On top of run budgets, each tenant carries rate limits and monthly quotas. Exceeding them returns distinct, documented error codes — RATE_LIMITED, QUOTA_EXCEEDED, or BUDGET_EXCEEDED — so workflows can tell a temporary throttle from an exhausted budget.
See where the money goes
Cost dashboards break usage down per run, per agent, and per tenant: tokens, model cost, latency. The platform also manages model tiering (fast models for classification, stronger models for generation) and prompt caching, which reduce cost without changing your agents.
Adjusting budgets
Budgets and quotas are governed settings. Run ceilings change through the agent's descriptor — a reviewed Catalog change; tenant quotas are managed with your NeoXam contact during the pre-GA programs.
Still stuck?
Tell us what you were trying to do and where it failed — screenshots and run IDs help. A human follows up.