To enhance the safety and adaptability of human-machine interaction, a novel variable stiffness robot joint prototype is proposed. This prototype employs permanent magnets and hollow electromagnetic coils to form the variable stiffness mechanism, enabling rapid adjustment of joint stiffness by manipulating the system’s magnetic flux. The stiffness model of the variable stiffness joint was successfully established based on molecular current method and virtual displacement method. Subsequently, controllers for coordinating joint position and stiffness, as well as a fuzzy adaptive controller, were developed. Experimental results indicate that increasing joint stiffness leads to improved accuracy in joint position and trajectory tracking, albeit with a decreasing trend in precision as motion frequency rises. The fuzzy adaptive controller outperforms traditional controllers in controlling the step response and trajectory tracking performance of the variable stiffness joint, exhibiting superior tracking accuracy and faster response speed.