I present HarmonAI, a visual analytics tool visualizing AI generated harmonizations designed to assist composers with harmonizing a given melody. Harmonization is a complex task, which has no unique correct solution, but various options of combinatorial complexity to choose from according to musical suitability. The goal of this work is to make generative AI for composition accessible to the target audience of non-professional musicians by communicating musical characteristics regarding harmony visually in a way that supports efficient browsing of numerous samples without the need of listening to them. Further, HarmonAI is meant to reduce the required music theory knowledge of the user to assess harmonizations and to allow estimating the effect created by a combination of melody and chord progression. My three central contributions are two visualizations and the extention of multiple metrics to give insight into progression of a piece of music: The (1) interactive tabular overview with incorporated visuals of all metrics forms the core of HarmonAI. The detailed (2) Harmonic Leadsheet is a piano roll based dense representation of symbolic music at the level of single notes interleaved with the (3) progression metrics, which consider computational harmonicity, distances and functional analysis. For evaluation I conducted three usage scenarios demonstrating the application of HarmonAI. Besides successfully giving this proof of concept I found that melody chord consonance is comprehensively captured while identifying potential for improvement regarding the visualization of characteristics related to chord internal structure. In addition, the Harmonic Leadsheet also was particularly useful for validating and understanding musical metrics.