Leaf litter habitats shelter a great variety of organisms, which play an important role in ecosystem dynamics. However, monitoring species in leaf litter is challenging, especially in highly diverse environments such as tropical forests, because individuals may easily camouflage themselves or hide in the litter layer. Identifying species based on environmental DNA (eDNA) would allow us to assess biodiversity in this microhabitat, without the need for direct observation of individuals. We applied eDNA metabarcoding to analyze large amounts of leaf litter (1 kg per sample) collected in the Brazilian Atlantic forest. We compared two DNA extraction methods, one total and one extracellular, and amplified a fragment of the mitochondrial 18S rRNA gene common to all eukaryotes, to assess the performance of eDNA from leaf litter samples in identifying different eukaryotic taxonomic groups. We also amplified two fragments of the mitochondrial 12S rRNA gene to specifically test the power of this approach for monitoring vertebrate species, with a focus on anurans. Most of the eukaryote sequence reads obtained were classified as Fungi, followed by Metazoa, and Viridiplantae. Most vertebrate sequences were assigned to Homo sapiens; only two sequences assigned to the genus Phyllomedusa and the species Euparkerella brasiliensis can be considered true detections of anurans in our eDNA samples. The detection of taxa varied depending on the DNA extraction method applied. Our results demonstrate that the analysis of eDNA from leaf litter samples has low power for monitoring vertebrate species and should be preferentially applied to describe active and abundant taxa in terrestrial communities, such as Fungi and invertebrates.