The treatment of Parkinson's Disease (PD) remains a major clinical challenge in modern medicine due to the lack of a definitive cure and the complexity of managing heterogeneous symptoms. Patients often require lifelong medication, necessitating physicians create individualized treatment plans to address diverse symptom profiles. However, traditional approaches to medication management are often limited by factors such as high costs and the limited number of patients a doctor can attend to. Although deep learning techniques are applied in PD treatment, they typically lack interpretability and are confined to numerical data inputs. In this study, we propose a novel framework that leverages large language models (LLMs) to design personalized treatment strategies for PD. Our approach integrates patient information in natural language form and external textual knowledge, such as medical guidelines, to generate personalized prescriptions. To enhance effectiveness, we use Monte Carlo Tree Search (MCTS) to refine strategies and establish a robust medication recommendation dataset. To enhance reliability and interpretability, we incorporate Retrieval-Augmented Generation (RAG) and Chain-of-Thought (CoT) reasoning within the LLM system, ensuring that each proposed strategy is accompanied by step-by-step explanations and references to similar historical cases. We validate our system using the Parkinson's Progression Marking Initiative (PPMI) dataset, featuring longitudinal patient records that ensure a statistically robust evaluation of our approach. Experimental results show that, compared to DL models, our method surpasses human-physician performance, reducing Unified Parkinson's Disease Rating Scale (UPDRS) scores by over 1.4 points on average. Furthermore, over 43% of patients achieve clinically significant improvements. A detailed case study underscores the flexibility of LLM in dynamically adjusting medication plans for patients at different disease stages, highlighting its potential to advance personalized PD management in real-world settings.