The modelling of social-ecological systems can provide useful insights into the interaction of social and environmental processes. However, quantitative social-ecological models should acknowledge the complexity and uncertainty of both underlying subsystems. For example, the agent-based models which are increasingly popular for groundwater studies can be made more realistic by incorporating geohydrological processes. Conversely, groundwater models can benefit from an agent-based depiction of the decision-making and feedbacks which drive groundwater exploitation. From this perspective, this work introduces a Python-based software architecture which couples the NetLogo agent-based platform with the MODFLOW/SEAWAT geohydrological modelling environment. This approach enables users to design agent-based models in NetLogo's user-friendly platform, while benefiting from the full capabilities of MODFLOW/SEAWAT. This workflow is illustrated for a simplified application of Aquifer Thermal Energy Storage (ATES).