The adaptive feature extraction capability of deep learning algorithms and the ability to collect data throughout the entire process of the Internet of Things provide technical support for efficient supervision of agricultural product supply chains. The research aims to build a safety supervision system for agricultural product supply chain through IoT technology, achieve food safety traceability, and improve the risk identification ability and supervision efficiency of agricultural product supply chain. The research adopts a hierarchical holographic modeling method to construct a risk indicator system, identifies agricultural product supply chain risks through backpropagation neural networks, and combines IoT technology for food safety traceability, thereby establishing a complete agricultural product supply chain safety supervision system. The results show that the average time to trace food safety issues before applying intelligent supervision is 15.9 days, while the average time to trace food safety issues after application is 4.2 minutes, with the maximum time spent being only 13.7 minutes, and it can effectively identify risks in the supply chain. The results indicate that the proposed agricultural product supply chain safety supervision system has improved the accuracy of risk identification in the agricultural product supply chain and achieved full traceability from production to consumption. The research results contribute to improving the quality and safety level of agricultural products and enhancing consumers’ trust in agricultural products.