<p>The process of vendor selection is a multi-dimensional and intricate task that requires analyzing the financial health, amount of risk, industry standards, and the mood in the marketplace. The use of traditional methods that consider individual data points or fixed metrics usually leads to less than optimal procurement decision-making. The current paper introduces a new Multi-Agent Large Language Model (LLM) framework that will improve the process of vendor evaluation based on the analysis of both structured and unstructured data. Its framework uses specialized agents that target different aspects of vendor evaluation: financial analysis, risk profiling, sentiment monitoring, and industry benchmarking. Structured indicators that financial agents consider include liquidity, profitability and solvency. Exposure to geopolitical, operational, and compliance risks is assessed by risk agents. Sentiment agents identify sentiment based on the news articles, reviews, and social media to determine the perception of the people and stakeholders. Agents in benchmarking compare vendors to industry standards to determine outliers and best-in-class vendors. The combination of these insights provides a data-driven yet context-aware procurement recommendation generated by the framework. The qualitative assessments are used to complement quantitative metrics, allowing procurement teams to make informed and holistic decisions. The system does not only help in identifying the most appropriate vendors but also assists in managing relationships with suppliers in the long term whereby the system alerts of any risk and performance deviations in real time. The method enables organizations to concentrate vendor selection in tandem with strategic objectives, reduce risk, and enhance the long-term value. The suggested multi-Agent LLM scheme is a major development in smart procurement systems, which enhance more resilient and value-oriented supply chains.</p>