Clustering is the process of classification of data in any of the emerging fields in present day scenario. In this paper we have proposed a framework for “multi-agent based clustering (MAS)” which is based on the working methodology of individual agents that initiate to dissolve a cluster. The major identification in this paper is once a cluster is created by an agent distinct types of agents will communicate and send messages for performing a task given to them, this process continues till the dissolving of cluster. An agent has the capability to either create a cluster or kill the cluster. And agents can even perform Inter cluster communication (ICC) between various clusters that are created as per requirement or necessity. We used K-means and KNN algorithms and shown that ICC can improve a clustering environment.