Message filtering is important for distributed multiagent systems, where a large number of dynamic agents participate in the system activity, but a typical agent is interested in only a very small dynamic subset of the other agents. The agent must be constantly informed on the status of this subset, and this is achieved by message passing between relevant agents. Message filtering is required to reduce the communications load on the system, which could be prohibitive if each agent must communicate with all others in order to obtain the information it needs. This paper deals with the case of a multiagent virtual environment, where each agent has a location in 2D space, and is interested in a small subset of the other agents, either those within a fixed range—as treated by previous authors, or the k other agents nearest to it—treated here for the first time. Furthermore, we treat the case of a fully distributed system, where no central server(s) are available to coordinate between the agents. The main challenge is then to design protocols that perform significant message filtering, yet enable each agent to maintain a consistent image of the other agents it is interested in. These protocols are useful in multi-agent games, simulations, and other virtual environments in which the geometric relationships between agents are important. They could also be useful for mobile-commerce and cellphone-based gaming applications.