Neurological rehabilitation is crucial for those with stroke or spinal cord injuries. These conditions often lead to upper extremity motor impairments. Such impairments make rehabilitation necessary for daily activities. Robotic rehabilitation devices come with engaging interfaces, showing promise in sustaining patient engagement. They offer encouragement for addressing upper extremity motor challenges. The contributions of this thesis lie in developing an MRI-compatible three-degree-of-freedom (3-DOF) robotic manipulandum, designed for enhancing neurological rehabilitation exercises that focus on the wrist and forearm. Seamlessly integrated with an existing rehabilitation interactive computer game, the robotic rehabilitation device can dynamically provide assistance or resistance based on the control mode during the game play. This robotic device prioritizes compatibility with functional magnetic resonance imaging (fMRI) as a core feature. fMRI enables brain scanning to provide insights into brain activity during rehabilitation exercises and potentially transform rehabilitation strategies. The device's 3D-printed mechanism, special sensors, and pneumatic actuators ensure MRI compatibility. The developed robotic device underwent trial runs by ten healthy subjects to demonstrate its viability and identify potential areas for improvement, to enhance its effectiveness and fulfil its purpose more effectively in the future. The findings highlighted that two assistance modes notably enhanced user performance, whereas two resistance modes posed significant challenges. This work paves the way for practical deployment in real-world neurological rehabilitation scenarios, offering personalized and effective strategies for sustained patient rehabilitation.