This paper examines how LLM-based agents extend beyond traditional LLMs to address complex software engineering challenges.

• Explores applications across the software lifecycle including requirements, design, coding, and maintenance
• Highlights how agents combine LLMs with decision-making capabilities to overcome standalone LLM limitations
• Identifies key challenges and future research directions toward AGI in software engineering
• Provides a comprehensive survey of current approaches and their practical applications

This research matters because it maps the pathway from current AI coding assistants to fully autonomous software engineering agents, offering a blueprint for the next generation of development tools.

From LLMs to LLM-based Agents for Software Engineering: A Survey of Current, Challenges and Future