In this research, we proposed concept "knowledge as media" to realize knowledge lifecycle for communication support on the network. In the traditional view of artificial intelligence, Knowledge exists within agents such as human and computer programs. On the contrary, we regard knowledge exists among agents, i. e., knowledge is media which enables communication among agents. Under this view, we investigated knowledge from the abstract way to the practical one. Ontology is the most abstract form of knowledge, knowledge for community and knowledge for environments are intermediate, and the embodied knowledge is the most practical one. We have three major results. One is community-support system called TelMeA in which interface agents as participants can form a field of communication. We tested this system by the psychological experiment and showed that using interface agents like this system is effective for communication. The second result is a system to capture personal knowledge. The system called MindHeap can gather knowledge through WWWW browsing. If a user can pick up sentences in WWWW, the system memorizes the sentences and their contexts. The sentences are connected to each other through their contexts in this system. The user can retrieve related memorized sentences automatically and even summarize WWW pages based on memorized information. The third result is embodiment-based learning, in which embodiment knowledge is captured through actions. The agent can learn how its body is by moving around and sensing the environment.