This paper is concerned with the set‐membership consensus of cooperation‐competition multi‐agent systems (CCMASs) with the round‐robin (RR) protocol. First, the relationship between cooperative and competitive agents is featured by the graph theory. Second, the RR protocol is used to relieve the burden of data transmission among agents due to the limited network bandwidth. Third, a novel concept of set‐membership consensus for CCMASs is presented to disclose the bound of consensus errors. In the framework of set‐membership filtering, a novel co‐design framework of set‐membership estimation and set‐membership consensus is provided to realize the desired cooperation‐competition consensus, where an ellipsoid is obtained to estimate the agent state. The controller and observer gains as well as the ellipsoid parameters are obtained by a recursive convex optimization algorithm. Finally, an illustrated example is given to demonstrate the effectiveness of the proposed method.