Abstract: Nowadays, research concerned about the upgrading of industrial clusters puts more and more emphasis on the digging of interior upgrading mechanism. Under this background, based on the cluster network theory, we analyzed the knowledge learning and upgrading mechanism amid clusters with the variables of network. Firstly, we studied the four different features of upgrading and the force for each concerned. Secondly, we analyzed the core of endogenous industrial clusters——knowledge learning and upgrading mechanism of the firm, the mechanism of multiple agents’ action on the upgrading of industrial, and the coupling mechanism of both firms and other multiple agents, then we proposed the research model. finally, the paper also investigated the validity and explanatory power of the model. The research imply that under the support of the endogenous industrial cluster multi-agent knowledge network, the relationship between multiple agents is in favor of the implementation of interactive learning and exploration learning. Moreover, it will benefit the knowledge-innovation and ultimately affect the updating of industrial clusters.