As mammalian diversity declines, effective biodiversity monitoring tools are urgently needed. To address this issue, we assessed the detectability of mammalian environmental DNA (eDNA) by applying eDNA metabarcoding to stemflow samples. We collected stemflow samples six times from June 2024 to January 2025 from 11 trees in the Oji Zoo, Kobe, Japan. One tree was inside the cage and thus in direct contact with the captive mammals, whereas the others were outside the cages, i.e., could not be directly contacted by captive mammals. We amplified eDNA extracted from the stemflow samples with MiMammal primers targeting the mitochondrial 12S rRNA gene, and the PCR products were then sequenced. We detected eDNA of 22 mammal species, 14 of which were zoo species and eight of which were not. Environmental DNA of Japanese squirrels (Sciurus lis), which could directly contact the target tree, was detected at all six sampling time points. In addition, we found a significant negative relationship between eDNA-based detection and the distance from each mammal cage to each tree, as well as significant differences in the composition of the detected species between the trees inside and outside the cages. In conclusion, our pioneering analysis shows that stemflow eDNA metabarcoding can be useful for detecting and monitoring mammals. The potential for its application in natural settings will need to be explored in subsequent studies.