Omnifact
AI platform from Frankfurt combining several language models and internal company knowledge, with a privacy filter for regulated industries.
What it does
Omnifact brings models from OpenAI, Anthropic, Google and Mistral together in one chat interface. In so-called Spaces you upload internal documents or connect sources such as SharePoint, Confluence or GitLab; answers cite the passage they draw on. A Privacy Filter detects personal and confidential details and replaces them with placeholders before a request reaches an external model. Admins decide per group which models and hosting regions are allowed. Custom connections run via REST APIs and MCP.
Who it suits
Public authorities, administrations and companies in regulated industries that want centrally managed AI access for many employees, combined with access to company knowledge.
Data protection for businesses
The platform runs on Microsoft Azure in Frankfurt; you can restrict usage to models operated in the EU. The data processing agreement under Art. 28 GDPR is concluded electronically together with the terms when you sign up. According to the vendor, prompts, documents and outputs are not used for training, and contractual zero-data-retention agreements apply with the model providers. You can switch each model provider on or off individually; there is no DPA for Groq. Some sub-processors are based in the USA, such as Anthropic, OpenAI and services for email delivery, web search and product analytics; the DPA does not mention Standard Contractual Clauses. Backups are kept for 30 days, logs for at least 90 days. Enterprise additionally offers on-premises operation and your own models.
Limits
No free plan. Pro is capped at 50 users and 25,000 pages in Spaces, and premium models draw on a shared credit pool of €5 per user a month that expires at the end of each month. SSO is only available in Enterprise.