Within a complex and volatile construction context, a real time safety monitoring method featured by solving management retard problem is in need for effective prevention and control of onsite accidents. This paper aims on analyzing critical safety factors performed by hand-held construction tools and onsite labor behaviors, establishing man-tool safety management indicators system, and using neural network model to assess safety indicators (except the Boolean-type indicators). Meanwhile based on onsite safety inspection procedures, the neural network safety decision-making model is integrated with RFID technology so that five subsystems (respectively man-tool identification subsystem, safety status information transmission subsystem, real time inspection subsystem, information management subsystem and safety alert subsystem) constitute a complete Man-Tool Monitoring and Management System, as to ensure accuracy and efficiency of onsite safety management.