Complex logistics—such as parcel delivery, curbside recycling, and emergency response—depend on Vehicle Routing Problem (VRP) solvers that can handle multiple stops, depots, and tight time windows. Most open solutions run outside the spatial database, forcing data to shuttle through files and scripts. vrpRouting eliminates that bottleneck. Distributed as a PostgreSQL extension on top of PostGIS and pgRouting, it embeds optimization algorithms—including tabu-search and other meta-heuristics—directly in database-native functions. Users store streets, depots, and orders where they already live in the database and receive optimized LINESTRING routes and per-stop attributes with no ETL overhead. Under the hood, vrpRouting streams network geometry to its engines, respects turn restrictions and one-way rules from PostGIS, and solves city-scale instances under tight time frames. A single query can pivot from capacitated one-depot scenarios to multi-depot pickup-and-delivery or strict time-window problems by toggling function parameters. We showcase this capability using OSM data for Auckland, routing parcel stops, scores of vehicles, and depot-specific time windows across the entire urban road graph, then visualizing driver sequences live in QGIS through a direct PostGIS connection. The presentation will unpack the extension’s architecture and current function set and walk through the Auckland case study. By keeping optimization within PostgreSQL, vrpRouting enables practitioners to transition from raw road graphs to production-ready routes with minimal database calls, thereby preserving the open-source ethos at the heart of FOSS4G.