Stroke patients often have upper limb dysfunction, and the wrist, critical for dexterity and daily life, needs targeted rehabilitation. Robot-assisted rehabilitation outperforms traditional therapist-led training as it avoids reliance on therapists’ experience/skills, offers personalized programs per patient, and logs full-cycle data to evaluate rehabilitation. Yet existing robots focus on arm movement, ignoring the wrist’s unique physiological and motor traits, and lack full-stage integration in structure, control, and interface. To solve these issues, this paper proposes the NPU-Wrist, a wrist rehabilitation robot integrating human factors and ergonomics, with a robot structure, control methods and interactive system. It adjusts to users’ hand, wrist, and forearm size, offering targeted training for patients in different rehabilitation stages (early, middle, late). A pilot study evaluated its usability, and results confirm its good ergonomics, comfort, and safety, satisfying personalized wrist rehabilitation needs.