In this work I present a tool which enables interaction for cluster-flow parallel coordinate plots. The goal of my implementation is to provide a user with ability to explore relationships between dimensions of a multivariate data set by rearranging the axis order and highlighting data points through brushing. The approach is evaluated by analyzing data related to a certain classification tree and attempting to find correlations between the visual results and the tree model. I discuss general problems of data representation in parallel coordinate plots and compare several techniques of clutter reduction. Further, I present the findings related to user experience during the interaction and propose potential solutions for detected problems.