Exploring intelligent mobile agent (MA) technology for assisting medical services or transmitting personal patient-health information in telemedicine applications has been widely investigated. Conversely, peer-to-peer (P2P) networking has become one of the most popular applications used in the Internet because of its benefits for easy-to-manage resources and because it balances workloads. Therefore, constructing an agent-based telemedicine platform based on P2P networking architecture is necessary. The main purpose of this paper is to construct a safe agent-based telemedicine that based on P2P networking architecture. Two themes are addressed in this paper: (a) the P2P network architecture for an agent-based telemedicine service, and (b) the security mechanisms for the proposed telemedicine networking architecture. When an MA contains patient information and migrates from one host to another through the Internet, it can be attacked by other software agents or agent platforms that can illegally access patient information. The proposed P2P network architecture is based on the JXTA protocol and provides two types of telemedicine service models: the predictable service model and unpredictable service model. This architecture employs a two-layer safety mechanism for MAs (i.e., time-limited black boxes and RSA undetachable signature technologies), to provide a secure solution for agent-based telemedicine services.