Abstract: The integration of artificial intelligence into emergency knowledge services has significantly enhanced the intelligence of knowledge organization. This paper proposes a multi-agent–based operational architecture for emergency knowledge services. Through the logical decomposition of individual agents, three core modules—data integration, knowledge reasoning, and knowledge fusion—are designed to enable standardized processing and semantic interoperability of multi-source heterogeneous data. By constructing a multi-agent task framework, the behavioral patterns and interaction mechanisms of decision-making agents, rescue agents, and public agents are defined, facilitating the establishment of a collaborative service system. The proposed architecture operates through a knowledge-flow-driven and parameter-matrix-coupled mechanism, supporting a continuous cycle of knowledge generation, fusion, and application in dynamic environments. This framework enhances the efficiency of collaborative decision-making and improves the accuracy of knowledge services during emergency response, thereby providing both a theoretical foundation and an operational framework for the development of intelligent emergency management systems.