Year: 2025
Type: conference-paper
Abstract: Software development involves complexities in requirements analysis, code generation, and continuous validation. This work presents a multi-agent system (MAS) designed to enhance the software engineering process by integrating Large Language Model (LLM) capabilities with human oversight. The MAS comprises five specialized agents that communicate through a shared message broker and maintain context in a persistent knowledge base. These agents autonomously handle tasks such as architectural planning, artifact generation, code review, and operational management, while human experts intervene to ensure accuracy and strategic alignment. The findings indicate that LLMs can impact software engineering, offering tools and techniques that automate tasks, enhance code generation, and potentially improve the development process.
Authors: Juliana Gomes Ribeiro, Joel Machado Pires, Jonas Oliveira Pereira, Frederico Araujo Durão, Célia Ghedini Ralha
Institution: Universidade Federal da Bahia
Language: English
Subfield: Information Systems
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7093104683