On-premise, made in Germany

Local AI for your company

Ziya ONE on ZiyaOS, powered by local AI.

Maximum security: the complete AI platform, on your own hardware, in your own server room. 100% sovereign, on German soil.

What happens in your company stays in your company.

Outside
  • Cloud AI
  • Licence server
  • Remote access
Your firewall0 bits out
Your server roomRunning
  1. Ziya ONEand your apps
  2. runs on
    ZiyaOSConnectors, governance, runtime
  3. runs on
    Local AIOpen models, your hardware

For companies where nothing may go to the cloud

You have always kept your data in-house, and that is how it should stay. Patient records, network plans, design data, client files: some information does not belong on someone else's servers, however good the contract. With Ziya ONE on local AI your employees still get capable AI. Not as a stripped-down island solution, but as a complete platform behind your firewall.

Built for sensitive industries

  • Healthcare

    Hospitals, practices and care facilities that work with patient data.

    Patient records
  • Critical infrastructure

    Municipal utilities, energy, water and telecommunications, with the highest demands on security and availability.

    Network plans
  • Engineering and industry

    Design data, manufacturing knowledge and the know-how that is your competitive edge.

    Design data
  • Law, tax and finance

    Client and financial data that must not leave the building.

    Client files

Zero-bit policy

The Ziya zero-bit policy

When Ziya ONE and ZiyaOS run on local AI, not a single bit leaves your network. No prompt, no document, no answer, no usage statistics.

Outbound traffic

Prompts
0 bits
Documents
0 bits
Answers
0 bits
Usage statistics
0 bits
Total
0 bits
  • No connection to the outside

    There is no connection to a cloud AI, no licence check over the internet and no remote access you have not granted yourself.

  • Air-gapped if you want it

    If you want, Ziya runs the whole platform fully air-gapped, without any internet connection at all.

  • Updates as verified packages

    Updates and new models arrive as verified packages that your IT installs itself, whenever it chooses.

Out of the box, everything is closed. What you open is your decision alone.

Everything local. Not just the chat.

A local language model on its own is not an AI solution. Only with connections to your systems, with roles and permissions and with ready-made applications does it turn into work that actually gets done. That is why, with Ziya, the entire stack runs in your building: Ziya ONE runs on ZiyaOS, and ZiyaOS runs on local AI.

  1. 01
    What your employees use

    Ziya ONE and the apps

    Ziya ONE for every employee, plus apps for research, minutes, reports or proposals.

  2. runs on
    02
    The operating system underneath

    ZiyaOS

    Connectors, agent runtime, governance and model layer. This is where your systems are connected and your permissions are enforced.

  3. runs on
    03
    The foundation

    Local AI

    Capable open language models on your own hardware. Everything above runs its computations on it.

Stays with you

Your knowledge

Documents, search index and histories live on your servers and nowhere else.

Permissions and audit log

Who may see what, and what happened when, stays in your hands and in your audit log.

Others put a chat in your server room. Ziya puts an AI operating system in it.

From the hardware to operation, from one source

  1. 01

    Clarify the need

    Which use cases, how many users, which data.

  2. 02

    Size the hardware

    Matched to your use cases, not oversized. You buy the hardware yourself or through your IT partner, and Ziya tells you exactly which.

  3. 03

    Install

    Local AI, ZiyaOS and Ziya ONE in your infrastructure, behind your firewall.

  4. 04

    Connect

    Active Directory, ERP, DMS and file shares, with your existing permissions.

  5. 05

    Enable and operate

    Training for your teams through Ziya Campus, maintenance and model changes only after your approval.

From the first user

Ziya has no minimum number of users. Whether 50 or 5,000 employees: your own AI server pays off from the moment your data must not leave the building.

1505,000
Employees

German. All the way into the server room.

Ownership
German owners, no foreign parent company
Development
Built and supported by the Ziya team in Dortmund
Legal framework
German and European law
Hardware
Stands at your site

Frequently asked questions about local AI

Does all of it work without the internet?
Yes. If you want, everything runs fully air-gapped. Local AI, ZiyaOS and Ziya ONE work without any connection to the outside.
What hardware do we need?
That depends on your use cases and on the number of concurrent users. In the initial consultation the Ziya team clarifies what you need and names a concrete configuration, from a single AI server to a larger installation.
Are local models good enough?
Open models today are productive for research across documents, reports, minutes, summaries and many agent tasks. Ziya tests beforehand, on your own use cases, which model suits which task.
How do updates and new models get into the system?
As verified packages that your IT installs itself. Nothing is updated automatically from outside.
Can we hand individual tasks to the cloud later after all?
Only if you explicitly allow it. ZiyaOS is model-agnostic, so that is a setting rather than a rebuild. Out of the box, everything stays local.
Do we need our own AI specialists?
No. Installation, integration, training and support come from Ziya. Your IT keeps control without having to build an AI team of its own.
Outside
  • Cloud AI
  • Licence server
  • Remote access
Your firewall
Your server roomRunning
  1. Ziya ONEand your apps
  2. runs on
    ZiyaOSConnectors, governance, runtime
  3. runs on
    Local AIOpen models, your hardware

Talk to Ziya about your server room.

In a free initial consultation the Ziya team works out which use cases should run locally at your company and what hardware you need for them.