Metabarcoding has emerged as a robust method for understanding biodiversity patterns by retrieving environmental DNA (eDNA) directly from ecosystems. Its low cost and accessibility have extended its use across biological topics, from symbiosis to biogeography, and ecology. A successful metabarcoding application depends on accurate and comprehensive reference databases for proper taxonomic assignment. The 18S rRNA gene is the primary genetic marker used for general/broad eukaryotic metabarcoding due to its combination of conserved and hypervariable regions, and the availability of extensive taxonomically-informed reference databases like PR2 and SILVA. Despite its advantages, 18S rRNA has certain limitations at lower taxonomic levels, depending on the lineage. Alternative fast-evolving molecular markers, such as the mitochondrial cytochrome oxidase subunit I (COI) gene, have been adopted as widely used "barcoding genes" for eukaryotes due to their resolution to the species level. However, the COI gene lacks a curated taxonomically-informed database covering all eukaryotes, including protists, comparable to those available for 18S rRNA. To address this gap, we introduce eKOI, a curated COI gene database aimed at enhancing the taxonomic annotation and primer design for COI-based metabarcoding at the community level. This database integrates COI gene data from GenBank and mitochondrial genomes that are publicly available, followed by rigorous manual curation to eliminate redundancies and contaminants and to correct taxonomic annotations. We validate using the eKOI database for taxonomic annotation of protists by re-annotating several COI-based metabarcoding studies, revealing previously unidentified biodiversity. Phylogenetic analyses confirmed the accuracy of the taxonomic annotations, highlighting the potential of eKOI to uncover new biodiversity in various eukaryotic lineages.