In recent years, the agent construction system has introduced novel mechanisms for coopetition between Urban Investment and Construction Companies (UICCs) and Real Estate Enterprises (REEs) in China's real estate market. However, the dynamic evolutionary patterns of such interactions remain poorly understood. Under concurrent competition and cooperation, the interplay between UICCs and REEs has grown increasingly complex, involving multidimensional shifts in information transmission and mutual influence, which traditional methodologies struggle to comprehensively capture and analyze. This study constructs a dual-layer heterogeneous network model, integrating prospect theory (PT) and mental accounting (MA) with coopetition game theory, to rigorously analyze multilevel interdependencies between UICCs and REEs. The objective is to uncover evolutionary dynamics of coopetition under the agent construction system and accurately reflect multidimensional interactive behaviors among enterprises. Key findings include: (1) Synergistic enhancements in network connectivity and initial cooperation willingness significantly promote the deep diffusion of cooperative behaviors. However, REEs exhibit greater sensitivity to fluctuations in cooperation willingness compared to UICCs
(2) A nonlinear relationship between agency fees and project premium rates reveals REEs' "preference for independent development" under brand advantages, whereas UICCs demonstrate strong reliance on balanced equity allocation. These conclusions offer managerial implications for governments to refine agent construction policies and for enterprises to formulate dynamic coopetition strategies.