This paper focuses on agent-based approach to study the relationship between the individual behavior of participants and the overall development of a virtual community, to help people to better understand the interactive process, forecast and manage the community development. In this paper, we first analyze the strengths of agent-based approach and the features of virtual communities. We argue that agent-based modeling and simulation approach is well suited to study virtual communities. We then build an agent-based model for virtual learning communities (VLCs). Each participant in VLCs is modeled as an agent with cognitive and social characteristics. The interactive activities are decomposed into four modules: registration module, activation module, action module, and adaptation module. A discussion of general observations of the community behaviors and managerial implications is presented based on series of comparative simulations.