Know the cost before it runs. Know the value after.

Some work deserves the best model; most work doesn't. Pick economy, balanced, or quality per task and see the estimate before anything spends a cent. The dial sets the model policy underneath — you choose the trade-off, the platform executes it.

• — An ordered model chain per skill — if a provider degrades, the next one takes the request automatically.
• — A cost ceiling per run and a budget per agent per month, enforced before the spend, visible after.
• — Your own provider accounts and negotiated rates. Vocion adds no markup and takes no percentage of traffic.

Tokens are how vendors bill; outcomes are how you manage. Every run's cost lands on the agent, the workflow, and the result it produced — so the question "what does the support desk coworker cost us per resolved ticket?" has a number on screen.