StockSage: Multi-Agent LLM Powered Inventory Management System for Intelligent Supply Chain Optimization

作者：Pranita Pingale, Faiz Asif Shaikh, Om Sanjay Bhongale, Sanvesh Satish Patil · 发表于：International Journal of Innovative Science and Research Technology · 年份：2026 · DOI：10.38124/ijisrt/26mar1951

Modern supply chain management systems often suffer from fragmented decision-making where demand forecasting, inventory control, supplier management, and pricing operate as independent processes. This lack of coordination frequently leads to inefficiencies such as stockouts, excess inventory, revenue loss, and poor supplier utilization. To address these challenges, this paper presents StockSage, a multi-agent inventory management system powered by Large Language Models (LLMs). The proposed system employs four specialized agents responsible for forecasting demand, managing inventory levels, selecting optimal suppliers, and recommending pricing strategies. These agents collaborate through a structured two-round coordination protocol that enables cross-functional communication and adaptive decisionmaking. The system is implemented as a full-stack web application using modern technologies including Next.js, React, TypeScript, Prisma ORM, SQLite, and OpenAI GPT APIs. A Monte Carlo simulation framework is used to evaluate system performance against traditional baseline strategies such as static reorder policies, moving average forecasting, and fixed pricing methods. Experimental results indicate improvements in forecast accuracy, service level, inventory turnover, and revenue optimization. The results demonstrate the potential of coordinated multi-agent LLM systems to provide intelligent, explainable, and scalable decision support for modern inventory and supply chain management.