Nowadays the rehabilitation process involves the patient and the therapist, that must interact to recover the motion of limbs and the strength of related muscles to restore the initial functionalities. The therapy relies on the experience and sensitivity of the therapist that identifies the rehabilitation exercises which are necessary to recover the expected ability. To prevent inappropriate practices an interesting aid may come by mixing collaborative robots, namely Cobots, and additive manufacturing technologies. The proper integration of a Cobot assistant and custom-printed training objects enables a significant improvement in the effectiveness of the therapy action and the related user experience since the programmed trajectories can mimic the movements related to activities of daily living. To this aim, this work describes an integrated approach to support the design of Cobot assisted rehabilitative solutions. The object selected by the patient and therapist, the motion pattern, the clamping area, and loads on the limb represents the design requirements. The motion trajectories defining the specific training tasks are the starting point to the optimal placement within the Cobot workspace. Specifically, manipulability maps can provide an objective evaluation of the locations where the exercises are performed at the best of workspace and configuration of the Cobot. A simple upper limb rehabilitation exercise based on a demonstrative handle has been selected to prove the effectiveness of the proposed approach. The results confirm that the manipulability index can be adopted to drive the preliminary design of the Cobotic solution toward a feasible configuration.