New challenging deployment scenarios are integrating mobile devices with limited and heterogeneous capabilities that roam among wireless access lo- calities during service provisioning. This calls for novel middleware solutions not only to support different forms of mobility and connectivity in wired-wireless in- tegrated networks, but also to perform personalized service reconfigura- tion/adaptation depending on client characteristics and in response to changes of wireless access locality. The paper proposes the adoption of Mobile Agent (MA) proxies working at the wired-wireless network edges to support the personalized access of limited wireless clients to their needed resources on the fixed network. In particular, the paper focuses on how to predict device mobility between IEEE 802.11 cells in a portable lightweight way, with no need of external global posi- tioning systems. In fact, we claim that mobility prediction is crucial to maintain service continuity: MA-based proxies can migrate in advance to the wireless cells where mobile clients are going to reconnect to, in order to anticipate the local re- arrangement of personalized sessions. The paper proposes and evaluates different mobility prediction solutions based on either client-side received signal strength or Ekahau positioning, all integrated in the SOMA platform. Both simulation and experimental results show that SOMA can predict the next visited cell with a very limited overhead and enough in advance to maintain service continuity for a large class of wireless Internet services.