This paper describes a new framework using intelligent agents for pattern recognition. Based on Jordan Curve Theorem, a universal classification method called Hyper Surface Classifier (HSC) has been studied since 2002. We propose multi-agents based technology to realize the combination of Hyper Surface Classifiers. Agents can imitate human beings’ group decision to solve problems. We use two types of agents: the classifier training agent and the classifier combining agent. Each classifier training agent is responsible to read a vertical slice of the samples and train the local classifier, while the classifier combining agent is designed to combine the classification results of all the classifier training agents. The key of our method is that the sub-datasets for the classifier training agents are obtained by dividing the features rather than by dividing the sample set in distribution environment. Experimental results show that this method has a preferable performance on high dimensional datasets.