Modern learning platforms are modular ecosystems with decentralized storage of content, offerings, and metadata. LLM-based chatbots are increasingly integrated to support learners in their educational journey. This development highlights the importance of collecting and analyzing learning data to enable personalized and adaptive e-learning applications that improve outcomes across different offerings. Based on operational research [1] with the German Armed Forces (Bundeswehr, Bw), we identified the need for decentralized, middleware-based Learning Record Store (LRS) solutions that respect organizational policies for data governance, privacy, federated identities, and interoperability, using the Experience API (xAPI) as the standard for tracking learning experiences. Current literature reveals a significant gap in the LRS landscape. Existing approaches often lack mechanisms for effective data flow control, interoperability, and strict data separation in decentralized environments. They also fail to maintain optimized metadata for AI services. To address these shortcomings, this research examines the organizational and technical requirements for a decentralized, middleware-based LRS. Our contribution is a novel concept designed to function as a network of data enclaves, enabling seamless queries and semantic control across the organization. A key feature is the data flow control mechanism, which complies with the xAPI specification and supports functions such as selective data merging to manage access and data flow, optimizing data processing with LLM agents.