Model Vault with North

Model Vault can be used in two ways, regardless of vault type:

• Standalone: deploy models and call vault endpoints directly. Cohere manages the underlying inference infrastructure, including networking, serving components, and compute.
• With North: deploy and operate the North application and control plane in your environment (agents, tools, orchestration, and data), and configure North to use vault endpoints for model inference, with Cohere managing the inference infrastructure.

To use Model Vault with North, run the North application in your VPC/network and route model inference to your vault endpoints.

When creating a Vault, the Add North + option provisions a Vault with opinionated, North-appropriate defaults. As with Add Model +, you can select models and configure the minimum and maximum replicas.

• Creating a Vault
• Calling a Vault over the API