Preprint tests multi-agent AI system for manufacturing process planning

A preprint describes Design-to-Plan, a large-language-model multi-agent framework that converts 3D CAD models, 2D engineering drawings, materials and manufacturing rules into process plans. An orchestrator coordinates agents for feature recognition, drawing analysis, context fusion, retrieval, sequencing and tool selection alongside deterministic modules. Across 300 benchmark cases, the authors report 100% success for three downstream agents, Tool F1 scores of 95.9% to 97.6%, and a 60% to 68% reduction in token use for key planning tasks.