This paper describes recent results of a unified computerized system for hand diagnosis and rehabilitation. Automatic diagnosis data collection and Virtual Reality rehabilitation exercises are the main characteristics of the system. The diagnosis subsystem includes a tactile sensing glove in addition to standard devices such as electronic dynamometer, pinchmeter and goniometer. Three standard rehabilitation exercises were simulated in a Virtual Reality environment, using the WorldToolKit graphics library. The first two exercises (ball squeezing and DigiKey) allow measurement of finger forces exerted during the rehabilitation routine. The third exercise (Peg board) involves the patient's visual-motor coordination. The rehabilitation subsystem uses a VPL DataGlove retrofitted with Rutgers Master (RM-I) and its interface. The exercises involve manipulation of objects with different stiffnesses and geometry. Grasping forces were modeled and fed back using the Rutgers Master worn on patient's hand. Data is gathered in real time from both diagnosis and rehabilitation subsystems. Finger specific forces recorded during rehabilitation exercises allow better diagnosis of the patient impairment. An ORACLE database is used to store and manipulate patients' records. Proof of concept trials were performed in a clinical environment. Some results of patient records analysis are presented in this paper. A new version of the system using an RM II haptic interface is presently under consideration.