Proximity based community planning organises essential services so residents can reach them within short travel distances, but its implementation requires combining large scale mobility data, multimodal accessibility modelling and spatial optimisation, components typically fragmented across different tools. We propose an open geospatial framework that integrates these three strands into one reproducible FOSS workflow. The framework has three components. The first performs mobility based community detection on telecom trip networks. The second computes multimodal accessibility on a 250 m grid, combining OSRM road routing with r5py walking and public transport routing over OpenStreetMap and GTFS. The third allocates facilities by linear programming whose objective jointly considers equity and economic efficiency. We demonstrate the framework on Busan Metropolitan City, covering 16 districts plus Gijang County, 770 km², 12,474 grid cells and 4,119 facilities across 18 categories. Community detection returns 50 stable living area communities derived from accessibility and 50 derived from mobility, with generalised modularity 0.918 and 0.934, mean best IoU 0.502, ARI 0.547 and NMI 0.814. Applied to a sample sub region of five inner city districts with 820,344 residents, the LP places 15 new facility units across 9 candidate 250 m grid cells and lifts walking only coverage of the five most deficient service types by 12 to 20 percentage points each.