The importance of knowledge management is growing day by day. In current environment, handling and managing of fast growing knowledge is very difficult which results in consuming more time for knowledge serving and therefore degrading the overall productivity. Therefore decision of choosing Multi-Agent system for implementing enterprise Knowledge Management (KM) results in reducing the time overhead for serving the relevant knowledge to the end-user by automatic communication between the agents. The main aim of this paper is to propose knowledge management system architecture for multi agents that will be helpful in disseminating knowledge to public sector organizations in a better way and thus enhancing the productivity. The knowledge management system architecture described in this paper also provides scalability, reusability and supports system adaptability. The paper first reviews the literature on various KM frameworks and the various tools for implementing multi-agents and also proposes Multi Agent Enterprise KM System (MAEKMS) architecture.