Knowledge-Based Multi-Agent Framework for Automated Software Architecture Design

TL;DR

The paper tackles the high cost and variability of software architecture design by proposing a knowledge-based multi-agent framework (MAAD) that automates the process. MAAD deploys four LLM-driven agents—Analyst, Modeler, Designer, and Evaluator—collaborating under a structured workflow guided by knowledge extracted from existing designs, authoritative literature, and architecture experts. The approach yields automated generation of requirements-aligned architectural decisions, UML/diagram artifacts, and deployment plans, coupled with rigorous evaluation and iterative refinement. It also discusses challenges in explainability, trust, coordination, and scalability, and outlines directions such as retrieval-augmented generation and expert knowledge integration to enhance reliability and adaptability.

Abstract

Architecture design is a critical step in software development. However, creating a high-quality architecture is often costly due to the significant need for human expertise and manual effort. Recently, agents built upon Large Language Models (LLMs) have achieved remarkable success in various software engineering tasks. Despite this progress, the use of agents to automate the architecture design process remains largely unexplored. To address this gap, we envision a Knowledge-based Multi-Agent Architecture Design (MAAD) framework. MAAD uses agents to simulate human roles in the traditional software architecture design process, thereby automating the design process. To empower these agents, MAAD incorporates knowledge extracted from three key sources: 1) existing system designs, 2) authoritative literature, and 3) architecture experts. By envisioning the MAAD framework, we aim to advance the full automation of application-level system development.

Figures (2)

• Figure 1: The Overview of Knowledge-Based Multi-Agent Framework for Architecture Design
• Figure 2: An Example of Proposed Architecture Design Process