Osteosarcoma, a leading primary bone malignancy in children and adolescents, is associated with a poor prognosis and a low global fertility rate. A large language model-assisted phenolic network (LLMPN) platform is demonstrated that integrates the large language model (LLM) GPT-4 into the design of multifunctional metal-phenolic network materials. Fine-tuned GPT-4 identified gossypol as a phenolic compound with superior efficacy against osteosarcoma after evaluating across a library of 60 polyphenols based on the correlation between experimental anti-osteosarcoma activity and multiplexed chemical properties of polyphenols. Subsequently, gossypol is then self-assembled into Cu2+-gossypol nanocomplexes with a hyaluronic acid surface modification (CuGOS NPs). CuGOS NPs has demonstrated the ability to induce genetic alterations and cell death in osteosarcoma cells, offering significant therapeutic benefits for primary osteosarcoma tumors and reducing metastasis without adverse effects on major organs or the genital system. This work presents an LLM-driven approach for engineering metal-organic nanoplatform and broadening applications by harnessing the capabilities of LLMs, thereby improving the feasibility and efficiency of research activities.