In today’s rapidly evolving world of technology, the need to modernize legacy software systems has never been more pressing. Organizations across different industries are struggling with challenges caused by outdated software. Legacy systems can often hinder innovation, reduce efficiency, and hide potential risks of failure. Modernization gives organizations the ability to prevent these difficulties and enhance organizational competitiveness, and adaptability in the ever-changing business environment. Software modernization, particularly the automated transformation of legacy code, is a critical task in maintaining the longevity and relevance of software systems. LLMs have shown promise in code transformation tasks, offering natural language understanding that can aid in code migration, refactoring, and documentation. Prior work in this area has focused on developing prompt-based prototypes for LLMs that guide the models through software modernization tasks, demonstrating LLMs’ ability to translate between languages, rewrite code, and suggest optimizations. While recent research has examined prompt engineering as a method to increase LLM accuracy in software-related tasks, there are limitations in prompt effectiveness across complex and diverse codebases. Agent-based architectures and RAG approaches, which augment LLMs with specialized agents or external knowledge sources, respectively, have shown potential for improving response consistency, accuracy, and contextual awareness. However, these approaches remain largely unexplored in the specific context of software modernization. The aim of this thesis is to evaluate and extend an existing Large Language Model (LLM)-based prototype for software modernization, focusing on optimizing prompt design for enhanced usability and effectiveness in code transformation tasks. In addition to refining prompt strategies, this thesis will explore agent-based and Retrieval-Augmented Generation (RAG) approaches to improve the adaptability and robustness of the software modernization process across diverse legacy codebases. By enhancing prompt design and testing more sophisticated LLM augmentation techniques, this research seeks to improve the practicality of LLMs in software modernization workflows.