Hexa-X will demonstrate the developed framework for the federated learning of XAI models, along with the related signalling, using a real-time network emulator and real terminals. End-user agents will sense local data and collaboratively train an XAI model with the goal to make predictions (e.g., on QoS) without exposing their data to other agents. The model obtained from the federated learning process will be transferred to other agents, which will adapt it locally to perform predictions. Moreover, an edge side-agent, federated with the end-user ones, and possibly with agents from other Mobile Network Operators (MNOs), will display in real time a dashboard, showing the predicted QoS in the federated domain and explaining how these have been achieved.