Abstract: The medical diagnosis system is a new complicated application system. The development of some new technologies such as Artificial Intelligent (AI), Computer Supported Cooperative Work (CSCW), and high bandwidth architecture made computer supported medical diagnosis available. Now the difficulty of the medical diagnosis is how to utilize the network which is the distributed processing platform to acquire the needed data and to reason in the condition of lacking the relative information. One of methods for solving the problem is to design an efficient architecture to integrate medical resources and expert knowledge to made them collaborative work. This paper describes a distributed medical diagnosis system using Multi-agent architecture and the middleware technology such as CORBA (Common Object Request Broker Architecture). The system could integrate more resources and facilitate collaboration among agents to minimize misunderstanding between patients and doctors. A diagnosis method based on FMMNN (Fuzzy Min-Max Neural Network) was applied to the architecture to verify the advantages of the system.