Cross-layering is a novel paradigm in wireless local area networks (WLAN) proposed to improve the support of QoS (Quality of Service) traffic. Many cross-layer solutions have already been developed. They diverge in the way the crosslayer concept is implemented and the type of performance gain they recommend. In parallel to cross-layering, a new network paradigm is growing: the autonomic paradigm appearing as the size and complexity of networks grow. There is a growing consensus among the research community that combining cross-layering and autonomic paradigms is, until now, the best alternative for better QoS support in WLANs. We combine cross-layering and autonomic concepts in the design of a new architecture called XLEngine (Cross-Layer Engine) having the ability to make optimizations based on local and network wide knowledge. XLEngine is implemented and simulated under the JSim simulator for different wireless scenarios. Simulation results show that XLEngine outperforms the classic layered architecture