AI model routing and data residency

Keydris Gateway lets your organization choose where a supported AI request is processed. You decide which workloads may use a cloud model and which must use a model on infrastructure you control.

Which model can see the data?

Consider a support team drafting replies. A question about a published return policy can go to Claude in an approved cloud deployment. A dispute containing a customer’s account history can go to a model the company hosts in its own data center. The team sets the permitted destination for each workload before using the gateway.

How does routing support data sovereignty?

Data residency depends on where the chosen deployment processes the request. For a workload that must stay in a particular region, the organization can approve a self-hosted deployment in that region. For a cloud model, the team needs to verify the provider’s processing locations and data-handling terms before approving that deployment.

What happens when an agent wants to act?

A team may let a model draft the reply and still require a supervisor to approve a refund. Routing determines where the draft is processed. When the agent requests the refund through a connected integration, Keydris agent authorization checks whether that action can proceed, needs approval, or must be rejected.

See the routing diagram or discuss a workload for the enterprise pilot.