We propose a new scalable architecture for mobile mixed reality (MR) systems and a content description language to be used in such architecture. Several architectures already exist to realize mobile MR systems, however, most of them are device specific. The architecture we propose here is able to accommodate a variety of devices, from mobile phones to notebook PCs. We have already designed a concrete specification for our architecture and content description language. We have also confirmed their viability by implementing several applications on various mobile devices.