With the development of precision, comprehensive, and full-course cancer care, multidisciplinary team (MDT) consultation has become an important model for promoting standardized cancer diagnosis and treatment and multidisciplinary collaboration. However, differences remain in the operational foundations and informatization levels of MDT across medical institutions, and practical problems include inadequate pre-meeting data preparation, insufficient integration of medical record information, inconsistent consultation documentation, incomplete tracking of post-meeting decision implementation, and non-uniform quality-control indicators. In recent years, large model and agent technologies can assist with case data organization, evidence retrieval, structured presentation of candidate diagnostic and treatment pathways, preliminary screening for clinical trials, follow-up reminders, and quality control; however, their medical reliability, data security, ethical compliance, human review, and boundaries of clinical responsibility still require standardization. To further standardize relevant applications, the expert panel reviewed domestic and international literature, existing oncology MDT expert consensuses, medical artificial intelligence (AI) application and evaluation consensuses, and clinical practice experience, and developed recommendations on scope of application, general principles, role settings, standard workflows, data governance, system evaluation, quality control, and responsibility boundaries. This consensus aims to provide a reference for qualified medical institutions to standardize the use of large model agent-assisted oncology MDT and to promote orderly implementation under the premises of patient safety, physician accountability, controllable processes, and continuous evaluation.