Abstract: By adopting local clustering, an automatic iterative clustering algorithm(LAICA) is proposed for complex networks. LAICA first agglomerates groups of nodes tightly connected as local clusters, and then iteratively merge local clusters based on some global modularity criterion. The experiment results demonstrate that LAICA can not only resolve the number of communities automatically, but also assign nodes into their communities accurately. The highest fraction of vertices classified correctly reaches 99.72% on real benchmark networks.