The management of the openness of a multi-agent system (MAS) is often done by a specific agent, called a middle-agent. The middle-agent knows representations of the agents and uses them to facilitate cooperations between them. But the centralization of these tasks has some drawbacks: there is a risk of overloading the middle-agent by several queries and the MAS can not support a middle-agent failure. In this paper, we propose an alternative approach to openness management which is completely distributed among the agents of the MAS.