Your knowledge stays yours.
Company knowledge can be processed locally and made available through your own knowledge and vector databases. Access rights and knowledge spaces remain under your control.
Use AI without giving up control.
Companies are increasingly building artificial intelligence into their core processes. That raises a new strategic question, and it is not a purely technical one.
- Who controls the models?
- Where is the data stored?
- Which information leaves the company?
- Which vendor does the infrastructure depend on?
- And what does AI cost once usage grows massively?
The Nebra Private Layer combines local AI models, company knowledge and intelligent agents into one infrastructure that you govern yourself. It gives you an additional option:
AI infrastructure under your own control.Your data.Your models.Your rules.
You choose the models.
Use, for example,
- Mistral
- Qwen
- other local models
and connect external models as well when needed.
You set the rules.
You decide which tasks are processed locally and when external AI is used. Not Nebra. Not a single model provider.
You.
Why should a company depend on a single AI?
Nebra does not try to replace Claude, OpenAI, Mistral or Qwen. Nebra creates the layer above them. Depending on the use case, you decide where a task runs.
Your company
Nebra Private Layer
Local
- sensitive data
- internal knowledge
- recurring processes
- private agents
Hybrid
- local data storage
- intelligent model selection
- external compute only where it makes sense
Cloud
- compute-intensive tasks
- specialized models
- tasks where external models are used deliberately
Use the best model for each task, but keep the decision in your own hands.
How a request flows through your AI infrastructure.
Every request passes through the Private Layer first. It checks your rules and sends the task where it belongs: to your own hardware or, if you allow it, to an external model.
Local + Cloud = Your choice.
Request
Nebra
Agents · Teams · Skills · Workflows
The workspace where employees and agents assign tasks.
Nebra Private Layer
Control based on your rules
OrchestrationPermissionsKnowledgeVector databasesModel routing
Local AI
Local models
MistralQwenmore models
Your infrastructure
Mac mini · Server · local systems
Cloud AI · optional
External models
ClaudeOpenAIGemini
Only for tasks your rules release, through your own contracts.
Local AI without external per-token API costs.
When tasks are processed entirely by local models, they incur no usage-based API token costs from external model providers. The computing power comes from your own infrastructure. You decide when local models are used and when external ones are.
For context: technically, local language models also work with tokens. The difference is who provides and bills the computing power. For local workloads, that is you, not an external provider charging per request.
Private AI should not be an IT project of its own.
You do not need an AI manager to run your own AI infrastructure. The Private Layer guides you through the necessary steps, and we handle the initial setup together with you.
- Choose a system
- Install local AI
- Select models
- Manage models
- Connect company knowledge
- Set up a vector database
- Define access rights
- Connect agents
- Orchestrate local and external
- Update systems
- Add new models
The complexity stays under the surface. What you get is a clear, understandable process.
Why private AI?
Data sovereignty
Sensitive information can be processed within your own infrastructure.
Model freedom
No dependence on a single AI provider.
Cost control
Local workloads require no external, usage-based API billing.
Knowledge ownership
Make your own company knowledge available to AI in a structured way.
Control
The company decides which systems may use which information.
Future-proofing
Integrate new models without rebuilding the entire infrastructure.
Your AI can live on your premises.
If you wish, you can source suitable hardware directly through us, such as an Apple Mac mini, and we will set it up together with you. Nebra does not depend on any particular device: the Private Layer runs just as well on a server or on systems you already operate.
- Mac minisourced and set up through us, optional
- Your own serveron-site or in the data center of your choice
- Existing systemsif they meet the requirements of the models
- NoteHardware is not part of the software and is offered separately.
Sovereignty starts with transparency.
No black box. What runs where, and who decides what:
Which components can run locally?
The language models (for example Mistral or Qwen), the knowledge and vector database, the agents and the control layer itself. They run on hardware at your company or on a server of your choosing.
When are external APIs used?
Only when your rules allow it for a given task. Without approval, a task stays local. You use external models such as Claude, OpenAI or Gemini through your own contracts with each provider.
Where is data processed?
Local tasks are processed on your own infrastructure. If a rule releases a task to an external model, exactly the content required for that task goes to that provider. You define which knowledge areas are eligible in the first place.
How are models selected?
Per task or knowledge area: either assigned permanently or based on your policies, for example "confidential data always stays local, public research may go external." New models can be added without rebuilding the infrastructure.
How do permissions work?
Company knowledge is organized in knowledge spaces. You decide which people and which agents may access which space. An agent only sees what it has been granted access to.
Two layers, clearly separated.
Nebra is the workspace for your agents. The Private Layer is the infrastructure beneath it, for companies that want to bring AI more firmly under their own control.
The AI workspace
from €99per month plus VAT
- Agents
- Skills
- Teams
- Workflows
- Multi-model use
- AI workspace
Your own AI infrastructure
from €499per month plus VAT
Included or depending on the setup:
- Private Layer software
- Local model integration
- Orchestration
- Knowledge infrastructure
- Model management
- Ongoing development and updates
- Support
One-time setupapprox. €1,500 to €2,500
plus VAT, one-time
Hardware, Mac mini, custom infrastructureoptional
Pilot phase: we are currently setting up the Private Layer together with our first companies. We define scope and pricing after a conversation about your setup.
Technology creates possibilities.Strategy decides what comes of them.
In addition to the software, you can book strategic guidance. The AI Advisory Board is a separate consulting engagement and is not part of the Private Layer price.
AI Advisory Board
€2,500 to €5,000per month plus VAT, guide price
Depending on:
Schedule a conversation
Nebra Private Layer
Private AI.Your Rules.
Digital sovereignty does not mean doing without the best AI models in the world. It means deciding for yourself when and how you use them.
Your AI. Your data. Your infrastructure.