This paper describes VisAdj, a tool for visualizing graphs as adjacency matrices. Although unfamiliar, this technique displays each link without occlusion and lends itself well to direct manipulation. Row and column permutations are a necessary step towards revealing the underlying structure of a graph. Since individual manipulation of vertices would take too much time in large graphs, VisAdj makes it possible to compute cluster hierarchies and manipulate them interactively. Aggregated views of the graphs can be accessed through these clusters and provide better understanding of their structure.