This paper deals with the integration of multimedia conferencing facilities into an existing software environment. We assume that the "semantics" of the conferencing facilities have already been defined and address only the issue of how to implement those semantics in such a way as to minimize the impact on existing software. In particular, users should be able to invoke any application from within the framework of a conference, and programmers should be able to develop new applications without having to deal explicitly with eonfereneing issues, We present the basic architectural approachs to this problem and discuss an early experiment with one of them.