Design and Development of Multi-Agent Healthcare Assistance Framework Using LLM and RAG

Medical clinics field a steady stream of routine questions about reports, appointments, symptoms, and medicines. Answering them quickly and correctly takes staff time that is often scarce, and patients are left waiting or puzzled by medical jargon. This paper reviews how researchers have approached the problem and then describes a multi-agent assistant built on Large Language Models (LLMs) and Retrieval-Augmented Generation (RAG). The literature survey covers more than fifty studies on traditional clinic support, conversational health agents, biomedical language models, medical question answering, RAG, LLM-based agents, evaluation, and safety. Together they show that general-purpose LLMs know a good deal of medicine but still invent facts, that retrieval from trusted sources lowers that risk, and that dividing the work among specialised agents makes an assistant easier to control. Building on these findings, the proposed system places a central orchestrator in front of five agents: medical question answering, medical document analysis, appointment management, symptom checking, and medicine information. Each agent draws on approved clinic knowledge, and every reply passes a safety review before reaching the patient. Quality is to be measured with ROUGE, BLEU, semantic similarity, and LLM-based judging, together with clinician review. The assistant supports routine communication and does not replace professional medical consultation. Keywords—Medical clinics, large language models, retrieval-augmented generation, multi-agent AI, medical question answering, medical document analysis, symptom checker, medicine information, appointment management, healthcare AI