In traditional teaching, teachers and students conduct real-time teaching face-to-face, while the efficiency of teacher-student interaction is low. To address this issue, this article proposes an intelligent teaching interaction system based on MAS. We have conducted a demand and feasibility analysis of teaching interaction applications based on artificial intelligence, and established the basic architecture of the teaching interaction system. Then, the key interactive modules are implemented with the Multi Agent concept, and a communication model based on SOAP is implemented using Web Service technology. To meet the needs of the system, a grouping strategy based on learners’ learning objectives is also proposed to allocate personalized tasks to individuals with different knowledge sets. Finally, the composition, module division, and implementation method of a web-based and MAS based online teaching system are provided and tested. The testing results show that our scheme improves the openness and knowledge scalability of the system, achieves communication between multiple agents in the system, and provides learners with a more convenient and efficient interactive learning environment.