This paper presents a new multiscale pore network modelling framework for predicting saturated and unsaturated permeability of OPC-based cementitious materials using a novel algorithmic implementation. The framework fundamentally relies on the data on cement composition and current understanding of cement hydration kinetics and microstructural features. Central to the modelling framework is the ability to numerically estimate pore size distribution (PSD) from existing models and the ability to obtain snapshots of unsaturated microstructure for various degrees of saturation. The framework is an amalgamation of three important existing models: (i) particle packing model for predicting nanoscale PSD, (ii) cement hydration kinetics to estimate microscale PSD, and (iii) a pore network model to estimate the permeability. The proposed pore network modelling is validated against an extensive set of experimental data that includes a very wide range of materials. The predicted intrinsic permeability falls well within the accepted experimental range. Though fewer experimental data are available to compare, the predicted unsaturated permeability shows highly promising results.