This paper presents a distributed predictive control methodology for indoor thermal comfort that optimizes the consumption of a limited energy resource using a demand-side management approach. The building divisions are modeled using an electro-thermal modular scheme. For control purposes, this modular scheme allows an easy modeling of buildings with different plans where adjacent areas can thermally interact. The control objective of each subsystem is to minimize the energy cost while maintaining the indoor temperature in the selected comfort bounds. In a distributed coordinated environment, the control uses multiple dynamically coupled agents (one for each subsystem/zone) aiming to achieve satisfaction of available energy coupling constraints. The system is simulated with two zones in a distributed environment.