Abstract: Adaptivity is very often an important goal for software systems. This paper reviews existing approaches to achieving adaptivity in object oriented systems, particularly those using design patterns, and concludes that there are further opportunities for improving adaptivity in agent systems. The approach described proposes that agents should be coupled with the environment, rules and laws about agent behaviour being externalised in a continuously re-configurable knowledge repository. Tools have been implemented to support the re-configuration. Once new requirements are specified by business experts via the tools, the system automatically adapts its behaviour in the environment, without additional effort from developers. This novel approach pulls together a knowledgebase, configuration tools, and business experts as an integral environment through which the MAS achieves re-configurability ranging from overall infrastructure to individual policy sets. This fosters cost effective software evolution because much of the effort resulting from changes to business strategies and collaborations can be shifted from developers to customers, code change not being required since the environment maintains the dynamics instead.
