Choose where your agents run

AI Agent Location controls where PersonaStack runs the agent assigned to a provider connection. Choose a location that can reach the model endpoint and the services your task needs.

Before you start

Open the connection under AI Providers. If you plan to use your own infrastructure, first connect its Kubernetes cluster in the app. A saved provider cannot use a cluster that is missing from its available locations.

Hosted or your own cluster

PersonaStack Hosted: choose an offered hosted location to run the agent on PersonaStack’s infrastructure.
Your Kubernetes cluster: choose your connected cluster when the agent needs to run there, such as when it needs an endpoint on your local network.

The model service and agent location are different settings. Selecting your own cluster does not move a cloud provider’s model into that cluster. It controls where the agent makes requests and uses its tools.

1. Save the agent location

Complete the provider’s credential and model steps where required.
Open AI Agent Location.
Select the hosted location or connected cluster you intend to use.
Save the provider.
Open a dependent persona and check the location shown beneath its provider name.

The location belongs to the shared provider connection. To move just one persona, assign it a different connection with the intended location.

2. Connect an Ollama server

Ollama requires your own Kubernetes agent location. Select that location first. Enter an Ollama server URL that is reachable from the cluster, then open the model dropdown to verify the connection and choose a usable model.

A URL that works in your laptop’s browser may not be reachable from the cluster. Use the server address available on the cluster’s network. Complete optional authentication headers only when your Ollama endpoint requires them.

3. Route a private compatible endpoint

For OpenAI Compatible, enable Route provider traffic through a Kubernetes cluster when the endpoint is private or reachable only from that network. Choose the Route cluster. This mode runs the worker in the selected cluster so it can reach that endpoint there.

If the location is unavailable

Check that you are in the workspace containing the Kubernetes connection.
Open the cluster’s settings and read its connection status.
Check network reachability from the chosen cluster, including the endpoint hostname and port.
Read disabled capability messages in Persona Settings. Features such as LAN access and container builds depend on the selected execution location.

Once the provider is saved, choose the capabilities your persona needs and try a small task.

Need a hand? Contact the PersonaStack team.