ObjectiveTo design a "roly-poly" style interactive home-based hand function training device based on the principle of self-stabilization and recovery in light of the characteristics of post-stroke hand dysfunction and the current scarcity of home-based hand function training devices; and explore the feasibility and clinical value of this device in hand function rehabilitation training for stroke patients. MethodsThis study adopted a phased research design consisting of two stages: device development and clinical validation. At the first stage, focusing on the principle of the tumbler's automatic return to center, combined with the clinical needs of hand function rehabilitation training for stroke patients, the development of a home-based interactive hand function training device was completed, and the application and authorization of a utility model patent were obtained. At the second stage, 30 stroke patients with post-stroke hand dysfunction hospitalized at Beijing Bo'ai Hospital from May to October, 2025 were randomly divided into control group (n = 15) and experimental group (n = 15). Both groups received conventional rehabilitation therapy, the control group added routine hand function training every evening, and the experimental group added hand function training using the device developed in this study every evening, for four weeks. The grip strength, lateral pinch strength, finger pad pinch strength, three-finger pinch strength and wrist joint range of motion (dorsal-flexion, palm-flexion, radial deviation and ulnar deviation) were measured before and after intervention. The upper limb and hand motor function was assessed with Fugl-Meyer Assessment (FMA), and the modified Barthel Index (MBI) was used to evaluate activities of daily living (ADL). ResultsA self-stabilizing hemispherical hand functional rehabilitation device was developed and granted a utility model patent. After intervention, the grip strength, lateral pinch strength, finger pad pinch strength, three-finger pinch strength, wrist joint range of motion, and FMA and MBI scores significantly improved in both groups (|t| > 5.683, P < 0.05); except for the score of MBI (P > 0.05), all the other indicators were better in the experimental group than in the control group (t > 2.633, P < 0.05). ConclusionThe self-stabilizing, closed-loop home-based functional hand interaction training device designed in this study effectively improves hand grip, pinching function and wrist joint range of motion of stroke patients, significantly enhancing upper limb and hand motor function, and demonstrates feasibility for home-based hand rehabilitation training. However, the device does not provide notable additional improvements in patients' ADL, necessitating further optimization to enhance its clinical application value.