Zervanor — Models & Routing
Models & Routing

One model layer for every agent and conversation.

A unified catalog of cloud, platform and on-device models, named routers with automatic fallback, and a server-side gateway that holds your keys and meters usage.

Unified model catalog

Bring your own cloud keys, use platform defaults, or register on-device models — all in one org-scoped catalog.

Fallback routers

Name a router with an ordered list of targets; if one is rate-limited, failing, or offline, the next takes over automatically.

Gateway with key custody

An OpenAI-compatible gateway runs cloud calls server-side — your keys never reach the agent or device — and meters tokens per org and model.

Per-agent choice

Pick the model source per agent at install: your own key straight to the provider, or a router with fallback. Same selector for cloud and device agents.

Host-aware resolution

The same router resolves by host: a cloud agent goes through the gateway; a device tries its local model first, then falls back to the cloud through the gateway — keys never leave the platform. Within an organization, one device can even serve its on-device model to others.

Route every model through one layer.