Frequently Asked Questions

Yes, the HainzelProtocol is completely compatible with MCP and A2A and follows the standards.

Hainzelman offers "thinking AI agents"that act as digital team members and go far beyond conventional automation tools. Instead of merely delivering chat responses, the platform automates complete work steps and tackles complex, cross-system challenges:: 

• Orchestration: It coordinates processes across multiple departments and IT systems (e.g., ERP, CRM). 

• Data Utilization: Its strength lies in processing large amounts of unstructured data and broadly utilizing corporate knowledge as a foundation for sophisticated, automated workflows. 

The HainzelStack is an open and modular architecture framework for AI workflows. It breaks down the automated process into clearly defined, independent layers. 

This modularity secures unrestricted technological independence:: 

• Interchangeability: Building blocks can be exchanged or expanded at any time, whether it is an AI model, a cloud provider, or a database. 

• Investment Protection: The clear separation of the layers protects investments in workflows and agent logic, even if the underlying technologies change. 

The Hainzelman Platform is the immediately usable and secure reference implementation of the HainzelStack architecture framework. It delivers all components in a single solution. 

As a core component for a quick start, the platform offers apps like the Knowledge ExplorerThe Knowledge Explorer solves the problem of fragmented knowledge and inaccessible documentation by transforming company internal knowledge into a usable, secure, and intelligent resource. Employees can query policies, processes, or project details, and the AI finds and explains the relevant information from all documents. The app often serves as a central entry pointbecause it delivers measurable added value within a few weeks. 

The unique selling proposition lies in the unique combination of immediate productivity and guaranteed technological independence.Companies no longer have to choose between fast, but binding complete solutions from large providers (vendor lock-in) or complex, fragmented in-house developments. With Hainzelman, SMEs receive a ready-to-use, user-friendly platform for the rapid deployment of AI while simultaneously retaining full control and the freedom to exchange or expand components at any time. 

The platform is model-agnostic. The most important proprietary models (OpenAI, Anthropic, Google) and models hosted in the GDPR-compliant EU Cloud are available as standard. Furthermore, models hosted in private clouds (e.g., via Azure) and locally hosted or on-premise models can be integrated.

No. The corporate data provided are naturally not used for training the AI models.

The system uses a fine-grained role and rights model.This ensures that a user can never retrieve information via the platform that they would be denied access to in the source system (e.g., SharePoint). 

The Hainzelman Platform is ISO/IEC 27001 certified.Further measures include: 

• Isolation: Clear client and workflow isolation. 

• Encryption: End-to-end encryption in transit (TLS) and at rest (e.g., AES-256). 

• Access Control: Strict authentication and authorization (RBAC). 

Hainzelman is designed to be GDPR compliant.The platform allows operation either in European data centers or on-premises (in the company's own network).This gives companies full control over their sensitive data. It is ensured that no AI models are trained with the provided data. 

Hainzelman uses Retrieval-Augmented Generation (RAG) in combination with intelligent knowledge refinement.This refinement is a multi-stage process that includes, among other things: 

1. Pre-processing: Removal of irrelevant content (such as HTML code) from documents. 

2. OCR: Use of Optical Character Recognition for difficult formats like scanned PDFs to extract text. 

3. Post-processing: Automated agent workflows condense, aggregate, and structure raw data to create higher-value assets.

Hainzelman seamlessly integrates into existing processes and authorization models. Integration is done via standardized interfaces and connectors to common systems such as DMS (Document Management Systems), ERP (Enterprise Resource Planning), CRM (Customer Relationship Management), email, databases, cloud services, ticketing systems (Jira), and code repositories (GitHub).

• HainzelProtocol: Is the binding "rule of the game" or common "language" for the agents, which defines how tasks are handed over, results are reported back, and errors are handled. It enables the seamless cooperation of heterogeneous agents across organizational boundaries. 

• HainzelML: Is a declarative, human-readable language (YAML-based) that serves as a simple "construction manual" for workflows. It allows domain experts to describe workflows without needing programming knowledge..

The system is model-agnostic.It allows for the free choice of AI models used: 

• Proprietary models (e.g., OpenAI, Anthropic, Google). 

• Models hosted in a private cloud (e.g., Azure EU) or On-Premise. 

• Self-hosting of open-source models like Llama or Gemma is possible, ensuring that no data flows back to the original corporations.

The solution is universally applicable but is especially well-suited for: 

• Knowledge-intensive fields (Legal, Administration, Consulting). 

• Regulated sectors such as maritime industry, finance, and healthcare. 

• Companies with complex cross-system process orchestration (ERP, CRM) and a high risk of knowledge loss.

Hainzelman offers ready-to-use apps and agent teams based on the Workflow Engine: 

• Knowledge Management: Knowledge Explorer, Assistant (Chat). 

• Specialized Apps: Hainzelman Expert Companion for utilizing expert knowledge, apps for Contract Management (e.g., Contract Wizard), and certification processes. 

• Core Component: A scalable Workflow Engine.

Companies can start quickly and efficiently: 

• A GDPR-compliant AI chat solution and the Knowledge Explorer are available immediately . 

Thanks to the ready-to-use reference implementation and pre-configured "Hainzel Teams," standard processes (like document processing or customer communication) can be automated within days . 

• Customer-specific processes are usually implemented within a few weeks .

Customers have full control over the costs for AI models (tokens): 

1. Included Contingent: The license fee includes a predefined contingent of tokens. 

2. Customer API Key (BYO - Bring Your Own): The customer provides their own API key. The token costs are billed directly by the model provider (e.g., Azure), which offers full transparency and control over data policies (such as opting out of model training).

The business model is based on the combination of an immediately usable platform and a flexible architecture: 

• Platform as a SaaS Subscription. 

• Licenses for on-premise operation, optionally with support contracts. 

• Access to pre-built agent teams/apps. 

• Individual projects (Custom Agents, integrations, tools) and consulting.

The name alludes to the legend of the diligent, invisibly working helpers (Heinzelmännchen), modernized with the "AI" in the name. The platform follows this principle: AI that assists in the background—reliably and unobtrusively —while companies retain control over data and processes.

Hainzelman is a technology company from Halstenbek that develops practical, secure, and scalable AI products for European mid-sized companies. The team consists of experienced AI and enterprise architecture experts.