Skip to content

Free AI consultation.Book a call

KROTERION
Book a call
Book a free consultationBook a demo

Self hosted AI for businesses on servers in the EU.

Powerful open source language models on servers in the EU, via API or in our chat app.

Book an initial call

Free and without obligation. Or see the packages

Try it outClick throughTry it out

Your team
KROTERIONServer in the EU

No detour via big cloud providers

What self hosted AI brings your business.

With self hosted AI for businesses, your requests stay on servers in the EU that we operate. Your team works with powerful open source language models in our chat app, and your developers integrate the same models into their own applications via API. Costs stay predictable because we do not bill per request.

  • Self hosted in the EU

    The models run on servers in the EU that we operate ourselves. Your input never goes to the big cloud providers.

  • Via API or as an app

    Developers connect the models through an interface. Everyone else works in our chat app in the browser.

  • Predictable costs

    We do not bill per request. We agree on the scope together, matched to your team and your applications.

  • On your premises too

    If data must not leave the building at all, we install the language model on a machine at your site.

AI hosting in the EU, without a detour via big cloud providers.

With many AI services, your inputs end up in data centres of big providers, often outside the EU. With us they travel encrypted to servers in the EU that we operate ourselves, and they are not passed on. We also support you with questions about the EU AI Act and GDPR.

Typical cloud AI

Data path unclear

  1. Your team

    Asks a question about a contract

  2. Big cloud provider

    Data centres often outside the EU

  3. Unclear further use

    The provider can change its terms

Self hosted at KROTERION

Stays in the EU

  1. Your team

    Asks a question about a contract

  2. Encrypted connection

    Directly to servers in the EU that we operate

  3. Self hosted model

    The answer comes back, nothing is passed on

AI via API or as a chat app, for your team and your software.

Your team works in the chat app, an alternative to ChatGPT where your data stays on servers in the EU that we operate. Your developers integrate the same models into their own applications via API. Both run on the same servers.

For the whole team

The chat app as an alternative to ChatGPT

Open it in the browser and get started. Your team asks questions, uploads documents and has texts written, without any installation.

  • Runs in the browser on computer, tablet and phone
  • Upload documents and ask questions about them
  • Choose the model for each task
  • Individual accounts for everyone on the team

Chats from the team

  • SA

    Revise quote for key account

    Sales

  • PU

    Check framework agreement for deadlines

    Purchasing

  • CO

    Q3 revenue as a table

    Controlling

  • HR

    Plan onboarding in the warehouse

    HR

For developers

The AI API in OpenAI format

Our API uses the widespread OpenAI format. Many existing programs therefore keep running without changes, you only swap the address and key. Recurring workflows between your programs are handled by our AI automation.

  • Compatible with common AI libraries
  • Separate key for every application
  • Streaming responses if you wish
  • Predictable costs without billing per request

Example request

curl https://api.kroterion.com/v1/chat/completions \-H "Authorization: Bearer $KROTERION_API_KEY" \-H "Content-Type: application/json" \-d '{"model": "kroterion-standard","messages": [{ "role": "user", "content": "Summarise this contract." }]}'

Three open source language models for different tasks.

Each model has its strength, from the fast model for everyday office work to the model for code and data. You choose per task, in the chat app with one click and in the API via the model name.

  • Standard model

    For everyday office work

    Speed

    Depth

    Typical tasks

    • Answering emails
    • Summarising texts
    • Writing minutes

  • Large model

    For difficult questions

    Speed

    Depth

    Typical tasks

    • Reviewing contracts
    • Analysing long documents
    • Preparing decisions

  • Code model

    For development and data

    Speed

    Depth

    Typical tasks

    • Writing and reviewing code
    • Analysing spreadsheets
    • Explaining scripts

Predictable AI costs. Tailored to your team.

Many cloud services bill every request individually, and the bill rises the more your team uses the AI. With us, you do not pay per request. So you know in advance what to expect, even if your team uses the AI every day.

Billing
Not per request
Accounts
For everyone on the team
Costs
Predictable in advance

Always included

  • Operation on servers in the EU
  • Server updates and monitoring
  • New model versions as soon as they are tested
  • Personal help in German and English

Tell us how your team wants to use the AI. In the free initial consultation we find the right framework together.

Book an initial call

Requests to the AI

2.913 Anfragen

Monday to Friday
Mon
Tue
Wed
Thu
Fri
7101316
LessMore

Your costs this week

As agreed

Packages for your self hosted AI. Matched to your project.

We put your project together from these packages. You can start with one package or combine several.

  • AI models via API

    Access to our self hosted models through an OpenAI compatible interface, for your own applications.

    Ongoing accessGood starting point

  • Chat app for your team

    Our chat app with the self hosted models, in the browser and with a personal login for everyone.

    For your whole team

  • Local AI installation

    A language model on a machine at your premises. Your data never leaves the building.

    Project

  • Knowledge management with RAG

    Your company knowledge from manuals, contracts and minutes becomes searchable for employees and AI agents. Every answer cites its source.

    Project

  • Internal chat platform

    An AI chat for your entire workforce on your own server, with your rules.

    Project

  • Local speech AI

    Speech recognition and speech output on your own servers, for example for voice agents or dictating reports.

    For voice agents

  • Operation of local models

    We maintain, monitor and update your local AI server.

    Ongoing support

Tell us what you have in mind. In a free initial consultation we work out the right approach for your project together.

Book an initial call

Questions and answers about self hosted AI for businesses.

Your question is not listed? In the free initial consultation we clarify everything that matters for your use case.

Ask a question
Which open source language models do you use?

We use powerful open source language models, also called LLMs. You can choose between a standard model for everyday office work, a large model for difficult questions and a model for code and data. We regularly test newer versions and always tell you openly which model is currently running.

Are our inputs used to train the AI?

No. Your inputs are only processed for the answer and are not passed on to third parties. The models run on servers in the EU that we operate ourselves.

Can we connect existing programs to the AI API?

In many cases, yes. Our API uses the widespread OpenAI format. Programs that work with it usually only need a new address and a new key.

How are the costs for self hosted AI calculated?

There is no fixed price list, we clarify the right framework for your team together in a short conversation. Billing is not per request, so you know in advance what to expect.

Do we have to host and operate the AI ourselves?

No. We run the models on servers in the EU and take care of updates and monitoring. If data must not leave your premises at all, we install a language model directly on a machine at your site and, if you wish, take over operation there too.

Is the chat app a ChatGPT alternative for businesses?

Yes. Your team asks questions, uploads documents and has texts written, directly in the browser and without installation. Every person gets their own account. Requests run through our servers in the EU and do not go to big cloud providers.

Which hardware do we need for local AI?

That depends on how many people use the AI and which models should run. For individual teams a powerful mini PC is often enough, for the whole company rather a server with graphics cards. In our AI consulting we plan the right hardware with you and set it up if you wish.

How do we get started?

In the free initial consultation we clarify what your team wants to use the AI for. Then we set up your access. For the chat app you only need a browser, for the API a key that we hand over to you securely.

More services that fit well.

All AI services

Your next step: a free initial consultation.

Tell us which tasks take up a lot of time in your company. We will tell you honestly where AI pays off and where it does not.

  • Video call or on site
  • Free and without obligation
  • Servers in the EU
Book a free consultation

What happens next

  1. Send a request

    A few sentences about your plans are enough.

  2. Consultation

    We listen and give you an honest assessment of what is worth doing.

  3. Clear plan

    You receive a proposal and decide in your own time.