With the accelerating aging of the population, there is growing demand among elderly individuals for personalized and proactive health management. However, current systems face multiple challenges such as fragmented health data hindering comprehensive assessment and insufficient personalization to meet the diverse needs of the elderly. To address these challenges, an intelligent elderly healthcare system integrating LLMs with a multi-agent system is proposed to achieve personalized health management. The system constructs a unified knowledge base by integrating Q&A on common geriatric diseases, drug databases, and medical literature databases on geriatric diseases. It employs voice-dominant natural language interaction to guide elderly users in actively inputting data. An LLM optimized through prompt engineering serves as the core intelligent engine, enhancing both senior-friendly adaptation and comprehensibility of health consultation services. Two specialized agents—for medication guidance and health knowledge consultation—perform tasks such as drug-drug interaction detection and elderly health Q&A under the coordination of LLM. System efficacy was validated with Sensibleness-Specificity Assessment (SSA) scores evaluating the rationality and relevance of multi-turn health consultations and Precision and Recall metrics assessing medication guidance accuracy. Experimental results demonstrate that the system delivers professional yet comprehensible health recommendations during consultations while achieving high accuracy in critical healthcare services like medication management.