In the area of clustering, proposing or improving new algorithms represents a challenging task due to an already existing well-established list of algorithms and various implementations that allow rapid evaluation against tasks on publicly available datasets. In this work, we present an improved version of the MTree clustering algorithm that has been implemented within the Weka workbench. The algorithmic approach starts from classical metric spaces and integrates parametrized business logic for finding the optimal number of clusters, choosing the division policy and other characteristics. The result is a versatile data structure that may be used in the context of clustering for finding the optimal number of clusters, but mainly for loading datasets, which already have a known structure. Experimental results show the MTree manages to find the right structure in two clustering tasks, although other algorithms fail in various ways. A discussion of topics related to further improvements and experiments on real datasets and tasks is included.