Robotic testing offers potential advantages over conventional methods including coordinated control of multiple degrees of freedom (DOF) and enhanced fidelity that to date have not been fully utilized. Previous robotic efforts in spine biomechanics have largely been limited to pure displacement control methods and slow quasi-static hybrid control approaches incorporating only one motion segment unit (MSU). The ability to program and selectively direct single or multibody spinal end loads in real-time would represent a significant step forward in the application of robotic testing methods. The current paper describes the development of a custom programmable robotic testing system and application of a novel force control algorithm. A custom robotic testing system with a single 4 DOF serial manipulator was fabricated and assembled. Feedback via position encoders and a six-axis load sensor were established to develop, program, and evaluate control capabilities. A calibration correction scheme was employed to account for changes in load sensor orientation and determination of spinal loads. A real-time force control algorithm was implemented that employed a real-time trajectory path modification feature of the controller. Pilot tests applied 3 Nm pure bending moments to a human cadaveric C2–T1 specimen in flexion and extension to assess the ability to control spinal end loads, and to compare the resulting motion response to previously published data. Stable accurate position control was achieved to within ±2 times the encoder resolution for each axis. Stable control of spinal end body forces was maintained to within a maximum error of 6.3 N in flexion. Sagittal flexibility data recorded from rostral and caudally placed six-axis load sensors were in good agreement, indicating a pure moment loading condition. Individual MSU rotations were consistent with previously reported data from nonrobotic protocols. The force control algorithm required 5–10 path iterations before converging to programmed end body forces within a targeted tolerance. Commercially available components were integrated to create a fully programmable custom 4 DOF gantry robot. Individual actuator performance was assessed. A real-time force control algorithm based on trajectory path modification was developed and implemented. Within a reasonable number of programmed path iterations, good control of spinal end body forces and moments, as well as a motion response consistent with previous reported data, were obtained throughout a full physiologic flexion-extension range of motion in the human subaxial cervical spine.