The small geographical range of highly endemic species is an important factor to be considered in conservation initiatives because it can increase their risk of extinction, as well decrease their probability of discovery. In this study, we use environmental niche modeling (ENM) to investigate the distribution of Brachycephalus, an anuran genus characterized by microendemic species living mostly in montane habitats along the Brazilian Atlantic Rainforest. Given that traditional ENM is not possible in the case of Brachycephalus because most of its species have limited geographical ranges, we analyzed an ensemble dataset that combined records of most described species, as well as new species that are currently being described, while accounting for heterogeneity in their climatic niches. Niche heterogeneity was quantified by ordination of the bioclimatic variables associated with their occurrence records, followed by unguided clustering of the resulting ordination scores. Out of an initial dataset of 544 records, careful curation reduced it to 75 records of 24 species and 71 localities. Interestingly, the three major clusters of climatic niches found in Brachycephalus corresponded largely to the three previously recognized phylogenetic lineages in the genus. The pernix cluster included the highly endemic species from southern Brazil that were most restricted to high-elevation areas, whereas the didactylus cluster encompassed species with broader geographical ranges that extended into lowland regions of the Atlantic Rainforest. Finally, the ephippium cluster included species from southeastern Brazil with intermediate levels of endemism. The detection of several isolated locations with potentially suitable habitats indicate that the diversity of Brachycephalus could still be considerably underestimated.