This paper proposes an approach to developing agents that can be reused in a number of collaborative applications. It particularly concerns those applications that evolve rapidly over time and need continual reconfiguration. This particularly applies to knowledge intensive processes in virtual environments, which requires quick reconfiguration to make then sustainable. It bases the agents architecture on a collaborative metamodel, which provides the ontology for building reusable software agents within collaborative environments. The paper describes the collaborative metamodel, an example and a prototype implementation and ways to integrate the model with agent systems.