The goal of this project is to develop a means for individuals with stroke to practice arm movement therapy at home with remote monitoring. We previously developed a Web-based system for repetitive movement training (Java Therapy). This paper describes a new input device for the system that measures and assists in naturalistic arm movement, as well as software enhancements. The new input device is an instrumented, adult-sized version of Wilmington robotic exoskeleton (WREX), which is a five degrees-of-freedom orthosis that counterbalances the weight of the arm using elastic bands. To test the ability of the new device (Training-WREX or "T-WREX") to measure and assist in functional arm movements, we measured five chronic stroke subjects' movement ability while wearing the orthosis without gravity balance compared to wearing the orthosis with gravity balance. T-WREX's gravity balance function improved a clinical measure of arm movement (Fugl-Meyer Score), range of motion of reaching movements, and accuracy of drawing movements. Coupled with an enhanced version of Java Therapy, T-WREX will thus provide a means to assist functional arm movement training at home, either over the Web in real-time, or stand-alone with periodic communication with a remote site.