The regulatory approval process for the manufacture, export, and import of medical devices mandates a series of procedures determined by device classification. Among these, the clinical trial phase is pivotal for demon- strating the safety and efficacy of the medical device. In particular, the thorough preparation of a clinical trial protocol at the initial stage is essential for facilitating the overall procedures and plays a key role in minimizing unnecessary time and cost expenditures during subsequent regulatory processes. Consequently, effective protocol preparation is a critical component in medical device development. However, practitioners in the field often face significant chal- lenges due to the need to comply with evolving regulations and the limited accessibility of up-to-date clinical trial protocol examples. While various support services have been introduced to address these issues, they frequently fall short in providing immediate and practical assistance. This study developed and evaluated MEDIVA (MEDical Inves- tigation Validation Assistant), a Retrieval-Augmented Generation (RAG)-based Large Language Model (LLM) tool designed to support the drafting of clinical trial protocols. The comparative analysis of completions generated by MEDIVA and a standard LLM lacking RAG capabilities based on pre-selected 4 representative questions likely to be encountered during protocol development demonstrated that MEDIVA provides more precise, actionable guidance through factually accurate and hallucination-free responses. It also delivers specific and consistent responses grounded in the latest regulatory information. The developed system is expected to extend its utility beyond protocol drafting, ultimately strengthening regulatory compliance capabilities, which are essential for the advancement of the medical device industry.