Sovereign AI agent platform for Swiss companies

The project: AI agents with control over data and models

AIgent is a sovereign Swiss AI agent platform. It is designed for companies that want to connect their knowledge and existing tools to AI agents without giving up control over data location, models or access rights.

The engineering challenge goes beyond chat: agents should review documents, reconcile data, produce reports and carry out further steps through connected systems. Those workflows need to remain traceable and follow each company’s roles, approvals and data-protection requirements.

The solution: agent platform, skills and sovereign operation

The agent platform, skills architecture and data infrastructure work together in AIgent. The architecture connects these components into a controlled workflow.

The case study is also a concrete example of our work on AI agents and workflow automation: from secure tool connections through to verifiable execution in production.

Agents and skills: recurring workflows are implemented as tested skills. An agent plans the task, uses approved tools and delivers results or files in the same workspace.

Agent harness and durable agents: execution does not depend on an open browser. A persistent event log records progress, allowing users to reconnect to their session after a connection drops. The agent harness manages tool calls, file access and approvals, turning a language model into an agent whose work remains traceable.

Company knowledge and integrations: documents and connected systems provide the necessary context. Roles, groups and single sign-on restrict which people and agents can access each resource.

Model choice and traceability: the model is selected for the task and data class. Execution steps, tools and results remain visible, allowing companies to review the agent’s work.

Deployment: AIgent runs as a Swiss cloud service in Swiss data centres or on-premise on the company’s own infrastructure, including isolated environments. The platform uses Next.js, Python, PostgreSQL and Kubernetes as a modular, operable product stack.

The result: a platform opening in stages

AIgent brings agents, skills, company knowledge, model choice and controlled deployment together in one product. A company can begin with one clearly scoped use case, then extend the platform across further teams and workflows. Access is opening in stages; requirements and the appropriate deployment model are discussed first.

Frequently asked about AIgent

How is AIgent different from a company chatbot?

A chatbot primarily answers questions. AIgent agents can also use approved tools, carry out work and create files. Tested skills make recurring workflows reproducible, while each step remains traceable.

Where do the data and models run?

Companies choose between the Swiss cloud and operation on their own infrastructure. The model can also be selected for each agent and data class. Sensitive data can therefore be processed locally while external models are approved for other tasks.

Our article on sovereign AI inference in Switzerland explains the engineering decisions that determine whether data storage and model operation are genuinely sovereign.

How can a company learn more about AIgent?

You can explore the platform and its features at ai-gent.ch.