We consider environments, like private houses, in which smart devices equipped with limited communication and computation capabilities have to cooperate to self-configure their state in an energy-efficient manner, as to meet user-defined requirements. Such requirements are expressed as scene rules, configured by the user using an intuitive interface that connects conditions on sensors' and actuators' states and actions on actuators. We translate this smart environment configuration problem into a constraint optimization problem. As to install distributiveness, robustness, and openness, we solve it using distributed message-passing algorithms. We illustrate our approach through a running example, and evaluate the performances of the implemented protocols on a simulated realistic environment.