Currently, kidney stone type identification is largely carried out visually through endoscopic images, where the visual recognition of the type of kidney stone by urologists is highly operator dependent. As such, deep learning (DL) models have the potential to develop in-vivo kidney stone identification models during an ureteroscopy, allowing for a more automated procedure that would expedite and increase the efficiency of determining anti-recurrence treatment prescriptions. Specifically, the DL model explored incorporated prototypical parts (PPs) that encompassed small, representative image patches. The PPs had the ability to capture color, brightness, and texture– the same traits lab experts utilize in standard analysis. This allowed for the generation of local and global explanations in determining what features were of importance and where they were in the image, aiding physicians in understanding why the DL made the corresponding identification decision. When tested on images of six common kidney stone types, this AI model was able to have a high overall average classification accuracy of 90.37±0.6%, serving as both a more explainable and accurate method for future kidney stone type identification.