Tremor, characterized by involuntary and rhythmical movements, is the most common movement disorder. Tremors can have peripheral and central oscillatory components which properly assessed and separated may improve diagnostics. An MR-safe haptic wrist manipulator enables simultaneous measurement of proprioceptive reflexes (peripheral components) and brain activations (central components) through fMRI. For such a MR-safe manipulator, this study determined the design criteria, created the design and presented its prototype. The prototype is divided into an MR-safe and MR-unsafe part. The hydraulic MR-safe vane motor (end effector) and optical sensors in the MR-environment are connected to the MR-unsafe part in the control room via hydraulic tubes and optical fibers. As a result, the fMRI quality is ensured and the manipulator is suited for safe use in any MR-environment. During a test in an MR-scanner (AMC, Amsterdam) no distortions in the MR-image or sensor signal were observed. The vane motor has a range of motion of 134° with a designed torque delivery of 8 Nm at a 3 bar pressure difference. The achieved accuracy for the torque sensor is ±2% full scale (F.S. = 14.3 Nm) and for the absolute position sensor ±1% full scale (F.S. = 70=1.22 rad). The prototype vane motor had some leakage along the axis, this prevented the use of pre-pressure and consequently reduced the bandwidth and maximum torque. However the maximum achieved torque of 1.5 Nm was still enough to met the requirement of 1.2 Nm. A PI controller was used to control the system, however this controller could not cope with the inherent non-linearities of hydraulics over the intended bandwidth (20 Hz). Despite the fact that the controller was not optimal, typical responses were obtained with impedance FRFs of two inertial loads. For future research it is recommended to remedy the axis leakage and implement a model-based controller. The open-loop FRF of the commanded velocity to the measured vane velocity was determined to characterize the system. Even with leakage and without pre-pressure, the measured bandwidth (-3 dB) was approximately 5 Hz and the -180° phase is passed at 7 Hz. Although the bandwidth does not fulfill the requirement for proprioceptive reflex identification it does allow for various motor control experiments. With the recommendations carried out, it is expected that the next prototype will be suited for proprioceptive reflex identification during fMRI.