Purpose/Significance To propose the framework MT-ARAG of the agent retrieval-augmented generation (RAG) for music therapy, and to enhance the application value of music therapy in digital health management. Method/Process A system consisting of two types of agents is constructed. The literature agent is used for academic literature retrieval and sheme design in music therapy, while the playlist agent is used to read users’ music preferences and use the Spotify tool to achieve therapeutic music recommendations. Cross-source knowledge integration and therapeutic music recommendation are achieved through the agent RAG framework. Result/Conclusion In the simulated clinical scenario, the MT-ARAG system has demonstrated its feasibility and innovation in the field of music therapy, and has the potential value of being transformed into personalized digital music therapy plans.