The Internet changes our shopping style. With the growing popularity of Web browsers, electronic commerce (EC) has become a trend of next-generation shopping style. EC software applications are written as Web document control programs, which run on service providers. The techniques used including information retrieval, network communication, database management, communication security and others. Due to the huge volume of data transmitted on the Internet, and the number of electronic commerce shoppers, currently the Internet is overloaded on its limited communication bandwidth. Research contributions are proposed to overcome this problem. Mobile agents are computer programs that can be distributed across networks to run on a remote computer station. The technique can be used in distributed information retrieval which allows the computation load to be added to servers, but significantly reduces the traffic of network communication. Many articles indicate that this approach is a new direction to software engineering. However, it is hard to find a theoretical base of mobile agent computing and interaction over the Internet. We propose a graph-based model, with a simulation design, for the mobile agents, which evolve over the Internet. Based on the concepts of food web (or food chain), one of the natural laws that we may use besides neural networks and genetic algorithms, we define agent niche overlap graph and agent evolution states for the distributed computation of mobile agent evolution. The proposed model can be used to build an environment for many electronic commerce applications, such as advertisement agent or survey questionnaire agent.