The mammillary bodies are a central component of the hippocampal-diencephalic-cingulate network that is critically involved in emotion processing and memory function, yet their genetic architecture remains unknown. We conducted a genome-wide association study (GWAS) of mammillary body volume in 38,064 participants from the UK Biobank and replicated our findings in the Rhineland Study cohort (N = 4,860). Mammillary body and hypothalamic subregional volumes were derived from structural brain MRIs using FastSurfer-HypVINN, an established and validated deep learning-based segmentation tool. Univariate GWAS identified 34 genomic loci associated with mammillary body volume, including 26 novel loci, with 20 loci showing replication in the Rhineland Study. Meta-analysis across both cohorts (N = 42,924) identified 44 associated loci, of which 35 were novel. Multivariate GWAS of hypothalamic subregions identified 55 loci, including 40 novel associations. Integrative cross-omics analyses identified MEI1 and CAPZA1 as key molecular candidates, with genetically regulated MEI1 expression and circulating CAPZA1 protein levels partially mediating genetic effects on mammillary body volume. Genetic correlation and Mendelian randomization analyses further revealed a bidirectional relationship between mammillary body volume and major depressive disorder. Together, our findings elucidate the genetic architecture of mammillary body volume and its relation to mental disorders.