The mobile agent is a computer program that is able to migrate continuously among hosts in a network and use host service to meet its task. The host,known as workplace, can be regarded as a proxy of social member. The sequence of workplaces on which the mobile agent completed its tasks is called path. In this paper, we propose a dynamic building method of mobile agent path with minimum payment based on referral. By referral, the next workplace of mobile agent can be recommended by the current workplace provider based on his acquaintance knowledge. The simulation results on a random network model show that the more acquaintance relationships there are on the referral network, the more efficiently the mobile agent path can be built, and the fewer costs need to be paid on the path. The mobile agent is a computer program that is able to migrate continuously among hosts in a network and use host service to meet its task. The host, known as workplace, can be considered as a proxy of social member. The sequence of workplaces on which the mobile agent completed its tasks is called path. In this paper, we propose a dynamic building method of mobile agent path with minimum payment based on referral. By referral, the next workplace of mobile agent can be recommended by the current workplace provider based on his acquaintance knowledge. The simulation results on a random network model show that the more acquaintance relationships there are on the referral network, the more efficiently the mobile agent path can be built, and the fewer costs need to be paid on the path.