Objective To achieve the unification of unilateral limb movement imagination and the actual movement rehabilitation, a MI-BCI based online prototype system is designed to improve the rehabilitation performance for upper limb motor function.Methods The system consists of the real-time motor imagery electroencephalogram (MI-EEG) acquisition module based on g.MOBIlab, the online processing module, the mechanical arm and control module, etc.The real-time acquisition, ocular artifact removal, feature extraction and classification of MI-EEG are applicable by hybrid programming with MATLAB and C Language, and as well as multithreading technology.The control module is designed for mechanical arm based on a microprocessor “S3C2440A”.The request response communication protocol between control module and personal computer is defined to control the corresponding extension/flexion action of mechanical arm during motor imagery.Results The average recognition rate of this system is 76.75% with 5 subjects being evaluated.Conclusions This system has excellent adaptability and real-time performance,which lay a solid foundation for development of more natural and advantageous clinical application system for upper limb rehabilitation.