Metagenomic analysis of deeply sequenced, eukaryotic-dominant symbiotic communities can be difficult for many metagenomic workflows. Here, we present MAGUS, a bioinformatic toolkit that uses a suite of custom bioinformatic methods for iterative genome assembly and filtering of pan-domain communities, where eukaryotes, bacteria, viruses, and functionally annotated gene catalogs are resolved and analyzed over a series of interconnected, modular software components. We evaluated MAGUS using deeply sequenced (median depth: 579 million reads) ten samples of hard corals, soft corals, and hydrozoans, which comprise complex, eukaryote-dominated symbiotic communities. We successfully resolved phylogenetically comparable host (N = 10), algal (N = 6), bacterial (N = 55), and viral (N = 160,925) genomes, as well as a gene catalog comprising 15,369,684 non-redundant genes (7.6% functionally annotated). MAGUS is available on GitHub (https://github.com/two-frontiers-project/2FP_MAGUS/).