A new multi-agent method for converting natural language to SQL has been developed, achieving 78.1% semantic accuracy on the BIRD benchmark. This advancement addresses the gap in NL2SQL accuracy compared to human experts, potentially enhancing database query capabilities for enterprises.