Understanding how agricultural management shapes biodiversity-ecosystem function (BEF) relationships is essential for designing ecologically sustainable farming systems. Using a 15-year field experiment with two tillage regimes (conventional tillage and no-tillage) and six fertilization treatments (single or combined mineral and manure additions), we examined how management practices restructure soil multitrophic communities and how these changes influence ecosystem functioning in a dryland agroecosystem. We characterized soil communities of bacteria, fungi, protists, and metazoans, quantified multitrophic co-occurrence network complexity, and assessed two functional dimensions: soil nutrient cycling based on enzyme activities and nutrient transformation potentials, and crop production based on yield and grain quality. No-tillage increased soil organic carbon and fungal richness but reduced network complexity, whereas fertilization primarily reshaped community composition without substantially affecting overall richness or multitrophic network complexity. Functionally, no-tillage improved nutrient cycling but reduced crop productivity, reflecting a management-driven imbalance between ecological and productive functions. Across management regimes, alpha-diversity showed weak relationships with either functional dimension, while community composition and network complexity were consistently strong predictors, particularly for nutrient cycling. Associations between soil biota and ecosystem functions were strongly context-dependent, varying with management practices and weakening or becoming non-significant after accounting for soil physicochemical properties, indicating substantial mediation of BEF relationships by soil conditions. Overall, our findings show that agricultural practices regulate ecosystem functioning primarily by reshaping multitrophic community structure and multitrophic networks rather than by altering species richness. This work advances our understanding of BEF dynamics in agroecosystems and highlights the need to manage soil community composition and multitrophic organization when designing sustainable agricultural systems.