There is function rvm in package kernlab which currently supports only regression. You can call it also via package caret. In addition, there is package rvmbinary.
I agree with Matthias Kohl suggestion. 'kernlab' is a very good R package for RVM albeit currently only support regression. Beside easy to use, the kernlab seems take full advantage of multicore processors. When I tried it for satellite image interpolation, most the cores in my processor are utilized up to 100%, although I have not yet confirm it by reading their documentation.
By the way, I also compile two tutorials on RVM in mind map for easier understanding and concise summary. Hope this would be a help.