Microorganisms are critical indicators of soil quality due to their essential role in maintaining ecosystem services. However, anthropogenic activities can disrupt the vital metabolic functions of these microorganisms. Considering that soil biology is often underestimated and traditional assessment methods do not capture its complexity, molecular methods can be used to assess soil health more effectively. This study aimed to identify the changes in soil microbial diversity and activity under different cocoa agroforestry systems, specially focusing on taxa and functions associated to carbon and nitrogen cycling. Soils from three different cocoa agroforestry systems, including a newly established agroforestry with green fertilization (GF), rubber (Hevea brasiliensis)–cocoa intercropping (RC), and cocoa plantations under Cabruca (cultivated under the shave of native forest) (CAB) were analyzed and compared using metagenomic and metaproteomic approaches. Samples from surrounding native forest and pasture were used in the comparison, representing natural and anthropomorphic ecosystems. Metagenomic analysis revealed a significant increase in Proteobacteria and Basidiomycota and the genes associated with dissimilatory nitrate reduction in the RC and CAB areas. The green fertilization area showed increased nitrogen cycling activity, demonstrating the success of the practice. In addition, metaproteomic analyses detected enzymes such as dehydrogenases in RC and native forest soils, indicating higher metabolic activity in these soils. These findings underscore the importance of soil management strategies to enhance soil productivity, diversity, and overall soil health. Molecular tools are useful to demonstrate how changes in agricultural practices directly influence the microbial community, affecting soil health.