: Multi-Agent systems follow a highly abstract programming paradigm. of this abstraction level, behaviours of the participants are not always clearly reproducible for humans. In this paper, we present a framework that enables agents to do some introspection and describe themselves. Our approach uses Java annotations and JDI, the Java debugging architecture. Programmers enrich the source code with comments for the user; during runtime, explanations are given and inform about what is going on "inside" the agent or whole system. We adopt our framework to the A4 MAS for supply chain management.