Abstract: Music composition represents the creative side of humanity, and itself is a complex task that requires abilities to understand and generate information with long dependency and harmony constraints. Current LLMs often struggle with this task, sometimes generating poorly written music even when equipped with modern techniques like InContext-Learning and Chain-of-Thoughts. To further explore and enhance LLMs’ potential in music composition by leveraging their reasoning ability and the large knowledge base in music history and theory, we propose ComposerX 1 , an agent-based symbolic music generation framework.