Long-read sequencing data can provide increased taxonomic resolution and genomic linkage information that is otherwise impossible to obtain from short-read amplicon or shotgun sequencing. However, the use of long-read sequencing for environmental DNA (eDNA) analysis has thus far been limited by both the apparent rarity of long DNA molecules in eDNA samples and the lack of established bioinformatics workflows for long-read metabarcoding, particularly from mixed template samples. Here, we report nanopore sequencing of 8.0 – 9.5 kb mitochondrial fragments obtained from mesocosm and field eDNA samples, amplified with long-range PCR (LR-PCR) using primers designed to preferentially amplify teleost mitogenomes We recovered half-mitochondria from 13 fish species in field-collected samples (Puget Sound in Washington State, USA), as well as from approximately half of the fish species inhabiting the positive control mesocosm (the Seattle Aquarium). Among biological replicates, we observed consistent detection of read-abundant species, while there was greater stochasticity in the presence of rarer species. We demonstrate that long fragments can be obtained from standard eDNA samples and successfully amplified and sequenced to obtain species identifications despite higher error rates characteristic of nanopore sequencing. We present both laboratory methods and an accessible bioinformatic pipeline for obtaining and analyzing LR-PCR amplified fragments from eDNA, providing a framework for future long-read metabarcoding studies.