We consider interactions in multi-agent systems, where part of the agents interact with a system that can be controlled (a super-agent or the environment) inducing some action to the agents. Such models arise in finance, traffic, smart grid, and social networks, We model the agents’ decision process as a Markov Chain and the controlled system as a dynamical system. We formulate the problem of controlling the probability that a set of agents make a specific decision as a Model Predictive Control problem and also as an explicit feedback law when the superagent can be modeled as a linear system or with a stochastic scenario-based MPC when the nonlinearity and the logical part of the superagent cannot be neglected. Simulation results from different everyday life examples (a broker suggesting investments to clients, traffic control in a two-line highway, or a group of people intending to purchase a specific brand t-shirt) are considered. The results reveal the potential of the considered modeling and control framework for applications where the decisions of a set of agents are to be influenced to achieve a desired objective.