Abstract: Describes an approach to design the architecture of a multiagent based intelligent decision support system(IDSS), using MFCADS as an example. MFCADS adopts a uniform hierarchical agent's internal control architecture. Agents are built based on the BDI framework. The agent's planning forms a partial ordered plan tree using HTN method. The component based method and the agent based method are used respectively to manage different types of models. In this paper the formalized description of models and an approach for modeling are presented. The work described solves decision support problems by using the multi agent theory, which goes forward one step further in reducing the separation of theory and practice in the field of MAS, and points out a new way to improve the efficiency and reliability of IDSS.