Objective: The mobility of the wrist joint is critical to the accuracy and stability of hand manipulation. Individuals with movement disorders, such as stroke survivors, require repetitive rehabilitation to restore wrist function. Rigid exoskeleton rehabilitation robots are limited by various issues, including misalignment with anatomical joints and high inertia. Cable-driven robots, with their flexible structures, offer distinct advantages including reduced weight, improved human-robot interaction, and better joint alignment. Consequently, they mitigate the limitations of rigid exoskeleton systems. To facilitate accurate wrist rehabilitation, a 3 degrees of freedom cable-driven wrist rehabilitation robot (CDWRR) is proposed. Methods: An adaptive configuration design is developed for a 3 degrees of freedom cable-driven mechanism designed to meet the functional demands of wrist rehabilitation, with a focus on user comfort and wearability. The open-structure CDWRR utilizes the human skeleton as a support structure and models the wrist joint as a constrained hinge. A kinematic model is established, and both the inverse position and inverse velocity are derived. Subsequently, a static is constructed, and cable forces are optimized using quadratic programming to ensure positive, continuous tension within safe limits. The wrench-feasible workspace is determined by integrating cable force constraints with a boundary search method. Motion dexterity is evaluated using the condition number of the Jacobian matrix. Finally, an experimental platform is developed, and a passive compliant training strategy—combining passive motion and admittance control—is proposed for early-stage rehabilitation. The feasibility of the configuration and control algorithms is validated experimentally. Results: The proposed 3 degrees of freedom CDWRR achieved full radial/ulnar deviation and covered 87.7% and 38.7% of the activities of daily living motion range for extension/flexion and pronation/supination, respectively. Within the workspace, the condition number ranged from 1.6 to 4.0, indicating good dexterity. Under external torque, the robot's workspace shifted in the direction of the applied force. During passive compliant training, when the interaction torque exceeded a set threshold, the robot demonstrated compliant behavior to ensure user safety. The cable length and force remained continuous and stable throughout motion, with no significant fluctuations, confirming the system's operational stability. Conclusions: The proposed 3 degrees of freedom CDWRR incorporates an adaptive configuration design that offers lightweight construction, improved compliance, and high wearability. The experimental results demonstrate that the robot satisfies the key rehabilitation requirements for wrist range of motion and dexterity. The passive compliant training experiments validated the feasibility and applicability of the wrist rehabilitation mechanism, providing an effective solution for early-stage wrist joint rehabilitation training. This paper provides a reference for the design and motion performance analysis of the CDWRR. The research results of this paper can provide a reference for the configuration design and motion performance analysis of the cable-driven wrist rehabilitation robot.