Agroforestry, the integration of trees within farmland, represents a promising strategy for enhancing agroecosystem resilience and mitigating climate change. The potential of agroforestry systems (AFS) lies in the ability of trees to improve resource use by accessing both aboveground and belowground compartments that are inaccessible to crops and to contribute to soil rehabilitation through increased organic matter inputs. However, little is known about how different AFS characteristics such as system age or tree density or diversity affect soil health in temperate regions. Owing to Switzerland’s diversified and hilly landscapes, farmers design site adapted AFS, resulting in highly heterogeneous AFS. Tree strip orientation and width, cropping alley width and use differ between farms. The degree of system diversity also differs: some AFS rely on a single tree species, whereas others combine several species or production goals. Likewise, farmers cultivate multiple crops in rotation and may even grow different crops simultaneously in adjacent alleys of the same AFS. This considerable diversity makes studying and comparing AFS particularly challenging as well as the extrapolation of field level findings to broader contexts. Moreover, agroforestry is frequently implemented alongside other soil conservation practices, such as reduced tillage, cover cropping, or increased organic amendments that have substantial impact on soils. Developing a more comprehensive understanding of how AFS structural characteristics and management practices shape soil properties is therefore crucial for assessing their contribution to agroecosystem resilience, particularly within highly mechanised agricultural systems. Within a network of 33 agroforestry farms across Switzerland, we investigated how AFS structural characteristics and management practices shape soil health at the plot scale. Soil samples were collected in the tree strips and arable strips of each AFS, as well as in a tree-less control plot managed similarly by the same farmer. Soil health was assessed using physical, chemical, and biological indicators. Soil management was evaluated based on soil tillage intensity, soil cover, and carbon inputs, using indicators computed with the SoilMangeR R package. Statistical analyses were performed with general linear mixed models, and the effects of agroforestry characteristics and soil management practices were examined while accounting for pedoclimatic factors, such as soil texture, mean annual temperature, and precipitation, at each site. Preliminary results show significant soil organic carbon gains in tree strips compared to arable strips and tree less control plots. While soil physical parameters appear to respond only weakly to agroforestry at this stage, soil biological indicators (including soil respiration, microbial biomass, and functional microbial communities) are higher in both the tree and arable strip compartments of AFS than in the control plots. Further analyses are underway to assess the effects of soil tillage intensity, soil cover, and carbon inputs, as well as their interactions with agroforestry characteristics, on soil health. This research will help support farmers in optimizing AFS design and management to improve soil health. Altogether, this work sheds light on belowground interactions in agroforestry systems and represents an important step toward understanding their contribution to more resilient agricultural landscapes. Acknowledgements We would like to express our gratitude to Nina Henny, Terence Cigler and Loïc Salomon for their great support during the field and laboratory work.