Due　to　the　shortage　of　natural　sand,　manufactured　sand　concrete　(MSC)　is　widely　used　owing　to　its　sustainability　benefits,　but　optimizing　its　mix　proportion　design　remains　a　challenge.　In　this　paper,　the　least　paste　theory　is　adopted　to　optimize　the　cracking　resistance　of　MSC.　Artificial　neural　networks　are　employed　to　establish　nonlinear　relationships　between　the　mix　proportion　parameters　(namely　nominal　water-cement　ratio,　equivalent　water-cement　ratio,　fly　ash-binder　ratio,　slag-binder　ratio,　stone　powder-binder　ratio,　and　average　paste　thickness)　and　performance　indicators　(namely　slump,　28　d　compressive　strength,　and　28　d　chloride　ion　diffusion　coefficient).　This　study　integrates　artificial　neural　networks　and　the　harmony　search　algorithm　to　optimize　the　mix　proportion　of　manufactured　sand　concrete,　enhancing　cracking　resistance　while　minimizing　cost　and　carbon　footprint.　The　findings　contribute　to　the　advancement　of　mix　proportion　design　theory　and　promote　the　broader　application　of　MSC　in　engineering　projects.　©　2025　Techno-Press,　Ltd.