On-premise AI.
Your models, your infrastructure.
The same product as the Swiss cloud, entirely in your environment. AIgent installs via a Helm chart on your Kubernetes cluster, up to air-gapped.
Three ways, one product.
From ready-to-go operation in the Swiss cloud to a fully isolated installation in-house. You choose the depth of control your data demands.
Swiss cloud
Operation in Swiss data centers under Swiss law. The default, ready to go.
Dedicated instance
An instance reserved for you with isolated resources, still operated by us.
Your own infrastructure
Entirely on your Kubernetes cluster, up to air-gapped with no internet connection.
A deployment in four steps.
Clarify requirements
We review data classes, directory services, models, and network segmentation together.
Helm deployment
Installation via our Helm chart on your cluster, the same product as the Swiss cloud.
Integration
Connect Entra ID, single sign-on, and role synchronisation from your existing services.
Operation
Choose models, set rights, enable the audit trail. Updates even in air-gapped operation.
The same product, entirely at your site.
Via a Helm chart, AIgent runs on your own Kubernetes cluster, up to air-gapped. Neither data nor models leave your environment.
You connect your directory services with single sign-on, choose models per data class, and keep the complete audit trail, even with no internet connection.
When data must not leave the premises.
On-premise is the choice for the public sector, banks, and healthcare, where especially sensitive data must stay in your own environment. Because AIgent runs the same product on your infrastructure, you give up nothing: agents, tested skills, model choice, and the complete audit trail are built in even air-gapped.
The decisive difference from cloud AI by US providers: there is no US parent company that could be compelled to disclose. Data control rests entirely with you.
Common questions about on-premise AI.
- What are the hardware requirements?
- AIgent runs on a Kubernetes cluster. The exact resource needs depend on whether you run models locally and how many agents run in parallel. For connected models only, a moderate cluster size is enough; local model inference adds GPU nodes. We define the concrete sizing together in the requirements review.
- Which open-source models are supported?
- AIgent is model-independent and can run open models locally as well as connect international models. You choose per agent which model processes which data class. In air-gapped operation the local models run entirely in your environment, with no data leaving it.
- How do updates work in air-gapped operation?
- Updates are provided as versioned artifacts and applied into your isolated environment, without the cluster needing an internet connection. The system stays current while the isolation is preserved.
- How is on-premise licensed?
- On-premise is billed as an annual license per installation, not per seat. Price does not scale with the number of users. We align the frame for enterprise and on-premise operation individually to your scope.
Sovereign AI. For every discerning company.
On your use-case, in your environment, within a clearly defined scope. See measurable results within a few weeks.
Secure early access. No per-seat fees. No lock-in.