Max's Grab Viewport is just a screen capture tool, so we can't do this directly. However we can post-edit the image (in Photoshop for ex.) and remove the background, but some automation of this process is welcome and also recently requested in CGTalk forum.
The first attempt about was Kostadin Kotev's script (here), follows by Leslie Van den Broeck (here). Leslie just extent the first script with more features (additional renderer passes), nice but the base solution is the same. This can be optimized to run about twice more faster, but the base part (process the bitmap per pixel) will remain too slow.
We (me and Kostadin) discuss an better idea and will work on it soon (I hope), so if you interested stay tuned.