This paper builds a multi-agents model to study the impact of the agents’ spatial decisions to the urban residential space evolution form and land use efficiency in the process of urban extension which contains one environment agent and three micro intelligent land use agents: residents, property developers and urban government. By the method of PEAS, the model induces the interactions among the three intelligent agents on the basis of the recognitions of their behavior characteristics and concludes that urban residential lands are developed in a priority sequence which is decided by the three intelligent agents’ spatial decisions and interactions. First, residents will choose favorable residential locations according to the law of consumption utility maximization. Then, property developers will choose these locations which can bring the maximized benefits. Residents and property developers’spatial decisions represent market mechanism which indicates the self organization of urban residential space in a degree. Furthermore, urban government agents will adjust the residential land developing priority sequence based on the comprehensive consideration of the social and ecological land use utility. So by adjusting the land use and environmental protection policies of urban government agent the model sets three scenarios which respectively represent the compact, relaxed and controlled modes of urban residential space extension and gets the residential developing priority function in every scenario. By the function, the model can get the preview of the evolution of residential space in every set scenario and provide land use planning policy guidance for urban government in advance. Taking Wuchang and Hongshan districts in Wuhan as the experimental areas, the paper compares the land use structure and land use efficiency in the process of the residential space evolution from 1998 to 2008 among the three scenarios and the actual situation respectively. The comparison indicates that land new development holds a lager proportion in the evolution process of the experimental residential space from 1998 to 2008, that is to say the land redevelopment level of old urban in experimental districts is not enough. In fact the government of Wuhan city had focused on residential new development of the suburban fringe areas before 2004, but the emphasis has been transferred to the old city transformation and land redevelopment after 2004, the point verifies the validity of the model to a certain degree. Compared with the model’s simulating results under different scenarios, the factual residential space evolution always has intersections with the three simulating results respectively, which means urban government may adjust its land use policy, natural environmental protection policy and so on under the influence of macroscopic environment in different periods. This is just one of the characteristics of Chinese real estate market. Excepting this, from the model simulating results, urban residents have attached increasing importance to the rights and interests of themselves, at the same time because of the action of market economic mechanism, the property developer also paid more attention to the favorite choices of urban residents; and the urban government gave more attention to the public willingness and the growth of resident welfare as well.