The base of research works, on interface agents, is the client-server concept, where they regarded them as humans’ assistants in the use of software. This vision does not permit the foundation, in a coherent way, of the humansoftware cooperation. In this paper, we propose a cooperation-based approach; of interface agents’development for multi-agent systems. Abstractly,we consider the human as an agent H that assumes one or more roles. Conceptually, we associate to the human a software F that holds the H’s objective. F is a software agent qualified as interface agent. However, F has neither human’s intellectual abilities nor its resources’ exploitation. We supplement this lack by F-human interaction. The agent H will then be consisted of F completed by the human’s intellectual abilities and his resources’ exploitation. H cooperates with other software agents and the development of F will follow a multi-agent development methodology.