Indoor thermal comfort is the most commonly studied type of comfort in the literature. We can find works which try to predict the user's satisfaction or keep static conditions by means of black-box controllers. However, we propose a novel system which is capable of adapting to the user's thermal preferences without any prior knowledge, and measuring his comfort level by aggregating several thermal parameters into one single thermal index. This single value is used in a static set of fuzzy rules easily understood by the user, and the labels used in those rules are dynamically adapted to the estimated preferences of the user. Experimental simulation shows that our proposal is capable of learning on-line the optimal thermal feeling for the user, and anticipating the necessary actions to obtain such thermal comfort in an indoor environment. The ubiquitous nature of nodes in a Wireless Sensor Networks (WSN) opens up wide possibilities for combining its distributed sensing power with advanced adaptive real-time learning systems. Thus, this paper also presents a proposal of integration of a thermal comfort adaptive fuzzy system into Coral2K (R), a centralized platform for monitoring and control heterogeneous WSN networks. (c) 2012 Elsevier B.V. All rights reserved.