A new multi-agent framework for automated concrete barrier design has been developed, achieving over 98% accuracy and outperforming larger models while reducing computational costs. This innovation could enhance the accessibility of AI tools in structural engineering, streamlining compliance with safety regulations.