Marine sponges are a prolific source of biologically active small molecules, many of which originate from sponge-associated microbes. Identifying the producing microbes is a key challenge in developing sustainable routes for production and isolation of sponge-associated metabolites, and requires application of several computational tools. To facilitate these analyses, we developed MetaSing, a reproducible singularity-based pipeline for assembly, identification of high quality metagenome-assembled genomes (MAGs), and analysis biosynthetic gene clusters (BGCs) from metagenomic short read data. We apply this pipeline to metagenome datasets from 16 marine sponges collected from New Zealand, Tonga and the Mediterranean Sea. Our analysis yielded 643 MAGs representing 510 species. Of the 2,670 BGCs identified across all samples, 70.8% were linked to a MAG, enabling taxonomic characterisation. Further comparison of BGCs to those identified from previously sequenced microbes revealed high biosynthetic novelty in variety of underexplored phyla including Poribacteria, Acidobacteriota and Dadabacteria. Alongside the observation that each sample contains unique biosynthetic potential, this holds great promise for natural product discovery and for furthering the understanding of different sponge holobionts.