In this paper, the communication effort required in a multi-agent system (MAS) is minimized via an explicit optimization formulation. The paper considers a MAS of singleintegrator agents with bounded inputs and a time-invariant communication graph. A new model of discrete asynchronous communication and a distributed consensus protocol based on it, are proposed. The goal of the proposed protocol is to minimize the aggregate number of communication instants of all agents, required to steer the state trajectories inside a pres-specified bounded neighbourhood within a pre-specified time. Due to information structure imposed by the underlying communication graph, an individual agent does not know the global parameters in the MAS, which are required for the above-mentioned minimization. To counter this uncertainty, the worst-case realizations of the global parameters are considered, which lead to min-max type optimizations. The control rules in the proposed protocol are obtained as the closed form solutions of these optimization problems. Hence, the proposed protocol does not increase the burden of run-time computation making it suitable for timecritical applications.