To address the issues of insufficient coordination among multiple stakeholders in park-level microgrids, inadequate exploitation of the power support value of flexible loads, and the difficulty of traditional demand response models in balancing economy, flexibility, and user satisfaction, a multi-agent Stackelberg game-based optimal scheduling method considering both energy and power values is proposed. First, an operation model for the Energy Producer (EP) is established, incorporating electricity trading revenue, power support revenue, and fluctuation penalty costs associated with the external grid. Then, for flexible loads including heat pumps, water pumps, and electric vehicles, a Load Aggregator (LA) operation model is developed, considering energy shifting, bidirec-tional power support, and user satisfaction simultaneously. On this basis, flexible load responses are guided through electricity trading prices and bidirectional power support prices, thereby achieving multi-agent benefit coordination and social welfare op-timization. To solve the nonlinear problem, linearization techniques and a bisection iterative algorithm are employed to obtain the Stackelberg equilibrium. Simulation results demonstrate that the proposed method can effectively smooth power fluctuations, enhance power support capability, and realize fair profit allocation among multiple agents while maintaining user satisfaction. Compared with conventional schemes, the proposed strategy increases the EP’s profit by 1.14%, reduces the LA’s cost by 7.59%, and improves social welfare by 35.82%. The proposed strategy fully exploits the resource potential of multiple agents, thereby enhancing both economic performance and operational flexibility while ensuring secure system operation.