What is Kasimir?
An overview of Kasimir, the GDPR-compliant multi-LLM AI platform for German mid-market companies, its features and its hosting in Germany.
Kasimir is a GDPR-compliant AI platform for German mid-market companies. Employees chat with leading AI models, analyze documents, transcribe meetings, run multi-step workflows, and collaborate in shared projects — all through a single secure interface, operated in Germany. The guiding idea: sovereign AI without a trade-off between capability and data protection.
The core idea
For a mid-market company, adopting AI today often means facing a dilemma: the most powerful models run on US servers, many "GDPR-compliant" alternatives are just thin wrappers, and most tools are built for engineering teams — not for sales, HR, accounting, or management.
Kasimir closes exactly that gap. By default, every request runs on open-source models (Mistral, Llama, Gemma, DeepSeek) self-hosted in a German data center — your data never leaves the infrastructure. When a task needs a frontier model (GPT, Claude, Gemini, Perplexity), it is processed GDPR-compliantly within the EU region, contractually safeguarded the same way, and without your inputs being used for model training.
One platform, no trade-off. Open-source models from the German data center for everyday work, frontier models GDPR-compliant in the EU for the demanding cases — both in the same interface.
Who Kasimir is for
Kasimir is a platform for every employee, not just technical teams. Typical user groups:
Sales & customer service — drafting reply emails, preparing quotes, writing customer correspondence
HR & administration — job postings, internal announcements, document research
Accounting & controlling — summarizing reports, analyzing spreadsheets, clarifying technical questions
Project teams — shared workspaces with their own chats, documents, and meetings
Management — quick research, decision briefs, meeting minutes
What Kasimir can do
Kasimir bundles a broad range of capabilities into one interface. The most important at a glance:
Feature | Description |
|---|---|
Multi-LLM chat | Chat across multiple models; open source from the German data center by default, GPT, Claude, or Gemini GDPR-compliant in the EU on request |
Agents | Specialized assistants with their own prompt, model, knowledge base, and access to your integrations |
Workflows | Multi-step, form-based AI processes without code — typed inputs, sequential steps, repeatable routines |
Document analysis | Upload PDFs, Word, or Excel files and ask questions about them — with clickable citations from the original |
Knowledge folders | Connect folders from Microsoft 365 or Google Drive — automatically synced and searchable by the AI |
AI web search | Answers with current web sources (via GDPR-compliant SearXNG) — with citations |
Data analysis | Upload CSV or Excel files — Kasimir analyzes them with Python and produces charts |
Database agent | Connect your own database and query it in natural language — secure and strictly read-only |
Meetings | Upload recordings or record directly — with speaker recognition, minute templates, and export |
Texts | Refine, version, and export longer AI-generated content in an embedded editor as real DOCX |
Translator | Translate documents and communication into 30+ languages — with your own glossary |
Dictation & read-aloud | Dictate by microphone and have answers read aloud — work hands-free |
Projects | Shared workspaces for teams — with their own chats, documents, and meetings |
Library | Ready-made agent and workflow templates — one click to adopt and customize |
Chat as the starting point
Chat is the heart of the platform. You ask a question, optionally attach files, and switch models when needed. Through the Werkzeuge (Tools) menu you selectively enable capabilities like web search, data analysis, or the database agent. Longer texts don't appear inline in the chat but in the sidebar editor on the right, where you can edit them directly.
Agents and workflows
An agent is more than a chat: a persistent persona with its own system prompt, preferred model, dedicated knowledge base, and access to integrations such as Outlook, Jira, or a database. Workflows, by contrast, are form-based and multi-step — ideal for recurring processes that should always deliver the same result at the same quality, without the team having to write prompts.
You don't have to start from scratch. In the Bibliothek (Library) you'll find dozens of ready-made agent and workflow templates — one click clones them into your workspace, where you can adapt them.
Data protection and hosting
At Kasimir, data protection isn't a bolted-on feature but architecture. The core commitments:
Hosted in Germany — data and self-hosted models run exclusively in a German data center
No third-country transfer for self-hosted models; frontier models are also processed within the EU region
GDPR-compliant — data processing agreement, data minimization, and the right to erasure are contractually anchored
No training on your data — training opt-outs are implemented in the contracts with all model providers
Self-hosted embeddings (BAAI/bge-m3) for document analysis — no third-country transfer during knowledge processing
Why not just use ChatGPT? Public services like ChatGPT store data in the US and use inputs for training by default. For companies with GDPR requirements, that isn't enough. Kasimir defaults to self-hosted models in Germany — and processes frontier models only GDPR-compliantly within the EU.
ISO 27001: An information security management system to the international standard is in preparation; certification is planned.
Integrations
Agents in Kasimir don't work in isolation but directly inside your existing systems — using your existing access and GDPR-compliantly. Already connected:
Microsoft 365 — Outlook Mail & Calendar, OneDrive, SharePoint
Google Workspace — Drive, Calendar, Sheets, Docs, Slides, Tasks
Jira and Confluence
Microsoft Teams
SQL databases (read-only)
In addition, folders from Microsoft 365 or Google Drive can be synced as a knowledge base, so the AI can search their contents.
Availability
Kasimir is currently in a closed beta. You request access via Beta beantragen (Apply for beta) on the home page. Pilot contracts are agreed individually; detailed pricing will appear with the public launch.
Next steps: Start with a simple chat, upload a document for analysis, and try a template from the Bibliothek (Library) to quickly get a feel for the platform.