The recent advances in wireless networking technologies and the growing success of mobile computing devices, such as PDA, pocket PC and third generation mobile phone, are enabling new classes of applications that present challenging problems to designers. The intrinsic constraints from limited device resources and frequent mobile disconnections cause that wireless applications can not be working smoothly as those in the traditional wired internet. Context-aware caching is a suitable approach to improve the quality of services in wireless applications. In this paper, we describe a context model, by which all contexts is divided into three categories as sensed, defined and derived. For the context prediction and selection in caching, we propose a weight-based data hoarding algorithm to select or update cached data objects. Two applications are introduced throughout in this paper to illustrate the effectiveness of context-aware caching.