Tri-MARF, a novel tri-modal multi-agent framework, integrates 2D images, text, and 3D point clouds with specialized agents to enhance 3D object annotation, achieving superior accuracy, retrieval, and throughput.