The control system based on cloud computing (cloud control systems) includes controllers, actuators and sensors which exist as cloud nodes. Information is exchanged among cloud nodes via the network. To solve the problem that the cloud control system is difficult to observe, which results from the wide application of the cloud node distributed running and matrix type data, a cloud supervisory system is designed and implemented. In this paper, the client$ / $server architecture of this system is described. The stability and richly functional controls are integrated into the system as the API by reconstructing, which meet the requirement of customize supervisory interface and improve the reliability of systems. The system also supports tuning parameters online to achieve consecutive observation through the duplex channel based on the TCP$ / $IP protocol. The data of cloud nodes is obtained and displayed in real time through asynchronous communication. All the elements of the matrix which is a common data type in cloud control systems are supported to supervise. Finally, for the motor plant, the Internet cloud-based dynamic simulation and Intranet cloud-based experiment demonstrate the feasibility and practicability of the designed system, and further promote and improve the control system combined with cloud computing in practical application.