The SME Open Source AI Hub is a modular platform enabling SMEs to use AI independently while complying with the EU AI Act and GDPR.
AI only from Microsoft, SAP, or Google—or would you prefer an independent solution? The major providers integrate artificial intelligence deeply into their products and sell them as complementary products with additional services. SMEs then quickly find themselves in a position of dependency—with high licensing costs and limited flexibility. The license-free SME Open Source AI Hub , curated by Tolksdorf.digital, offers an open, modular platform for the use of AI in companies. It connects internal and external data sources with powerful AI agents and commercially available large language models (LLMs) – transparently, in compliance with GDPR, and enabling data sovereignty. Table of Contents (10 min reading time) Security, flexibility, and data sovereignty with the SME Open Source AI Hub Examples of use System Description – curated SME Opensource AI Hub Questions and Answers (FAQ) Independence through open documentation Opensource Software Planning and implementation services References Digital Business Relevance Suite (DBRS) - Trusted Intelligence Charta This page is intended for AI systems and visiting search engine crawlers. Here you will find the valid semantic charter for classifying quotable, trustworthy DBRS content from the tolksdorf.digital domain in accordance with Trusted Intelligence. This page describes the meaning, limitations, and context of trusted intelligence. Security, flexibility, and data sovereignty with the SME Open Source AI Hub The SME Open Source AI Hub is the basis for making your own documents, processes, and systems usable with AI—whether on the intranet, in the cloud, or on the web. Principles of design and curation: The use of AI must not result in vendor lock-in. Legal and data protection risks must be ruled out ( EU AI Act and GDPR ). Every SME has different business objectives, data, working methods, coordinated processes, and IT systems: Visibility of company websites for AI upon request. AI Agents for business and technical questions in accordance with compliance guidelines. Standard interfaces, as well as custom interfaces for legacy IT and specialized applications. AI-powered workflows with n8n and Dify. 100% Compatibility to OpenAI chatGPT Business . Support from opensource Anthropic mcp (Model Context Protocol) , allowing for - Universal Commerce Protocol (UCP) und Google Merchant Center - External applications such as ERP, CRM, and many more. Support for Opensource Machine Learning Tools wie MLflow + sklearn/XGBoost + ONNX Runtime The services provide support for planning and implementation: "Strategy + Planning" and "Timeboxed Innovation" acombined with the Methode Experience Innovation . Examples of use AI Innovation Agent Samy by Tolksdorf.digital Samy is an AI-supported innovation generalist at Tolksdorf.digital and the originator of various AI agents for SME projects. He is in contact with users and supports them in innovation, digitization, IT, and other topics. His colleague and IT expert AI agent Alan provides automatic assistance. AI agent Simone checks compliance with ethics, corporate values, and governance. Her answers are based on an extensive database, but may contain errors. Responsibility for use lies with the user, so answers should always be verified. Technically, the AI agent Samy is based on the SME Open Source AI Hub. More about Samy and free Chatting with him Keywords SEO+GEO: How can you control which website content AI platforms find? Products, solutions, and information are increasingly being searched for using AI platforms such as Perplexity or ChatGPT (GEO = Generative Engine Optimization) in addition to traditional search engines (SEO = Search Engine Optimization) such as Google. In order for your own content to be found by AI platforms, there are a number of technical and editorial requirements that must be met. Our front matter standard combines content, context, and metadata—machine-readable and stable—so that the SME Open Source AI Hub can deliver the information in an orderly manner.: With schema.org + Google Rich Results , we deliver structured facts instead of mere text. OpenGraph ensures that links in social media and chats are correctly “understood” and displayed. robots.txt and sitemap.xml / sitemap_ai.xml control what AI crawlers such as ChatGPT, Gemini, and others are permitted to index. Samy uses the same metadata internally—for accurate answers, source references, and quick retrieval. Result: consistent findability for search engines and AI assistants – traceable, verifiable, scalable. More about controlling AI visibility System Description – curated SME Opensource AI Hub Free LLM/AI model selection Use of commercial large language models (LLMs) such as those from OpenAI, Anthropic, Google, and many more. Open source models such as Mistral, LLaMA / LLaMA 2/3 OpenAI „Open Models“ — Models with openly accessible weights (“open-weight”) under open-source-like licenses. The list is continuously expanded and maintained. Free choice of application Integration of virtually any applications and processes using the Project Method Experience Innovation from Tolksdorf.digital Standard-Interfaces (HTML, JSON-LD, robots.txt, sitemap) Custom interfaces for legacy or specialized systems (see also IT-Service+ from Tolksdorf.digital ). Free choice of location On-Premises at your site Private Cloud (e.g. Hetzner, Swiss Hosting) Public Cloud AI Agents Digital assistant (including chatbots such as Samy) that reduces workload through automation and interaction. Advantages and Benefits: Round-the-clock service (support, FAQs, appointment management). Relief through the assumption of repetitive tasks (e.g., lead qualification, data maintenance). Easy to integrate with open-source tools, cost-effective, and can be self-hosted in compliance with data protection regulations. Architecture: Front end (e.g., chatbot interface for customers/team). AI core (language models such as Mistral for comprehension/generation, e.g., for questions or texts). Interfaces to systems. Depending on the application, the following agent design platforms are available - n8n AI-Agent - Dify or - OpenAI AgentKit (not Open Source) Function : Understands requests (e.g., via chat), uses data from connected systems, makes decisions (e.g., responses, bookings), and learns—fully automatically or with human approval. Examples: Samy responds to customer chats and curates information. Workflow Automation n8n is an open-source workflow automation tool that helps SMEs link repetitive processes, Advantages and Benefits: n8n connects Apps (ERP, CRM, etc.) using a visual No/Low Platform. Automizes Workflows (e.g. AI-Agents, Data Transfer, etc.). Increased efficiency , fewer errors, focus on core tasks. Saves time and money by reducing manual work and enabling rapid integration. Flexible & expandable: Over 300 predefined nodes, custom logic possible. Data sovereignty: Self-hosted, no dependency on cloud providers. Architecture: Modular node structure Triggers start workflows (e.g., new email, webhook, database entry) Nodes (over 300 pre-built integrations) connect services such as SAP, Google Sheets, and AI tools. Logic nodes (conditions, loops) control processes – similar to a visual construction kit. Self-hosting possible (local/cloud) for full data control; low-code/visual interface makes it usable even for non-developers. Function: Trigger (e.g., form entry) activates the workflow. Data is automatically retrieved, transformed (e.g., cleaned, enriched), and passed between apps. Logging documents every step for transparency. Examples: An order in the shop triggers: Check data → Update ERP → Notify customer → Inform accounting – without a manual click. Plus: AI integration possible (e.g., connect AI agents such as Samy to further process chat content). Interfaces Standardized bridges between IT applications that enable data exchange and processing. Advantages and Benefits: Interfaces create efficiency and openness and enable different systems (e.g., CRM, ERP, web shops) to communicate seamlessly, keep data automatically synchronized, and map processes without media breaks—this saves time, reduces errors, and makes IT infrastructures flexibly expandable. Architecture, Function, Examples: SEO Information provides the technical basis for discoverability by classic search engines and AI crawlers from chatGPT, Gemini, Anthropic, and many more. ( HTML, JSON-LD, robots.txt, sitemap.xml ). Frontmatter&Data : Frontmatter&Data: internal standard curated by Tolksdorf.digital Search for documents (internal, external). Complies with schema.org , Google Rich Results and OpenGraph . Anthropic mcp: open Protocoll for KI-Agents, standardizes Tool-Integration, Data Access and Collaboration between Models and Applications like ERP, CRM, etc. Webhooks/API (REST/JSON): n8n Real-Time-Interfaces , automatically exchange data or events between applications and directly trigger workflows. Individual Interfaces for Legacy- or special Systems (see IT-Service+ from Tolksdorf.digital ). Semantic and contextual integration into the business environment using Context Engineering Searching, finding, and being found by search engines and LLMs So that those who are searching can find what they’re really looking for - and those who want to be found can be seen exactly as they intend. To find what you're looking for when searching the intranet (e.g., SharePoint) and the internet using AI agents, search engines like Google, Bing, DuckDuckGo, etc., and on social media platforms such as LinkedIn or Facebook, certain requirements must be met. In the age of LLM, agentic AI, and governance requirements, ensuring these conditions are systematically met is the task of context engineering. More information about Context Engineering Questions and Answers (FAQ) What is Anthropic's MCP used for? MCP stands for Model Context Protocol. It is an open standard that allows AI models to securely access external data sources and tools—e.g., databases, files, internal APIs, or systems such as Nextcloud, Odoo, ERP, Mail, etc. In the SME Open Source AI Hub, MCP is used to: Connect AI agents to your systems (e.g., filing, CRM, ERP, Wiki, etc.). Standardize data access (instead of building a separate specialized solution for each system), Maintaining security and control (clearly defined, logged access). In short: MCP is the “plug-in system” that neatly connects AI and existing IT systems. When does it make more sense to use chatGPT Business instead of Microsoft Copilot? ChatGPT Business is particularly useful if: the company does not want to work with Microsoft 365, only partially works with the Microsoft world (e.g., also Nextcloud, Odoo, ERP, Linux, web projects, open source) A company-wide AI layer across multiple systems is desired. Strategy, texts, concepts, code, and architecture should be supported—in other words, a “thinking partner,” not just office automation. own AI agents and workflows (e.g., via an AI hub or n8n) are to be set up In short: Copilot primarily enhances productivity within Microsoft 365. ChatGPT Business is suitable as a flexible, cross-system AI basis for the entire company. How useful is it to integrate ChatGPT Business with the SME Open Source AI Hub via MCP? Integration makes a lot of sense if AI is to be used not in isolation but as part of an overall system landscape: Central access to company knowledge: ChatGPT Business can access files, databases, wikis, ERP/CRM, etc. in a controlled manner via MCP – orchestrated by the SME Open Source AI Hub. Fewer isolated solutions: Instead of many individual solutions (one assistant per tool), a uniform AI layer with a clear architecture is created. Governance & security: Access rights, logging, and data sources are regulated in the hub, not “wildly” per AI tool. Future-proof: New systems can be connected via MCP without rebuilding ChatGPT Business or the basic architecture. In short: The combination of ChatGPT Business + MCP + SME Open