Game-theoretical approach is applied for multiagent optimized system engineering. All agents are intellectual players interacting with each other to meet their goals. In particular, they must attain system equilibria resolving rationally conflict of interests. Game rules and agents’ autonomous behavior are depicted in the language of monoidal category of binary relations. Players have not opposite interests which are described with the help of preferences relations that might change. They are to be maximized in some classes of permissible strategies. The latter depends on agents’ communication structure which is also to be picked out by agents themselves. The problem compositionality is studied. It is the most important property of multiagent systems. Game resulting relation is constructed to search not only system equilibrium under goals uncertainty but optimal communication network. The latter notion is defined. In order to construct it, decentralized method is proposed based on thinking-consciousness laws application. Contributed approach to multiagent systems study meets up-to-date networks technologies.