The faithful numerical representation of the coupled dynamics between groundwater (GW), the unsaturated zone (UZ) and surface water (SW) remains one of the more persistent challenges of regional-scale integrated hydrological modelling. Physically-based models that solve the Richards equation in three dimensions provide a rigorous description of vertical fluxes, but their computational cost, sensitivity to parameter uncertainty, and tendency to over-emphasise capillary forces at the expense of preferential and non-equilibrium flow regimes limit their suitability as self-contained UZ modules for regional integrated hydrological models, ensemble simulations, and multi-decadal climate-impact studies. Conversely, the conceptual and water-balance bucket models routinely adopted at regional scales rarely retain enough vertical resolution to track the position of a moving groundwater head, to represent capillary rise from the GW into a depleted root zone, or to impose a physically consistent evapotranspiration (ET) stress that responds to the local soil-moisture state. This paper introduces GWSWEX (Groundwater–Surface Water EXchange), a vertically resolved process-based modelling package designed to occupy the middle ground between these two extremes and to act as a UZ coupler between an external GW model and an external SW model within an integrated modelling chain. GWSWEX represents each model element as a layered 1-D soil column, tracks GW elevation, per-layer UZ storage, and SW ponding as prognostic states, and exposes two interchangeable numerical solvers behind a unified Python API: an explicit operator-split bucket-sequence solver with CFL-adaptive sub-stepping, and an implicit mixed-form Richards solver with Picard linearisation and Thomas-algorithm tridiagonal inversion. Both solvers share the same spatial discretisation, the same Mualem–van Genuchten retention and conductivity relations, the same vertically integrated drainable-volume function for the moving groundwater head, and the same atmospheric and lateral boundary conditions. The model is implemented as a modern Fortran 2008 kernel with OpenMP parallelism over elements, exposed to Python via f2py, and configured through a Pydantic-validated user-facing API. Verification against HYDRUS-1D across six soil profiles and two contrasting forcing scenarios, together with an iterated one-at-a-time sensitivity analysis of the empirical parameters of both solvers and a parallel-ensemble computational performance benchmark, demonstrates sub-centimetre to near-centimetre GWH accuracy across smooth and intensive forcing regimes, characterises the physical limits of the layered-bucket abstraction, and examines the computational cost of the model at the ensemble sizes that regional integrated modelling requires.