Abstract: Large Language Model (LLM) based multi-agent systems (MAS) have shown promise in tackling complex tasks, but often rely on predefined roles and centralized coordination, limiting their adaptability to evolving challenges. This paper introduces $\textbf{MorphAgent}$, a novel $\textbf{Autonomous}$, $\textbf{Self-Organizing}$, and $\textbf{Self-Adaptive Multi-Agent System}$ for $\textit{decentralized}$ agent collaboration that enables agents to $\textit{dynamically evolve their roles and capabilities}$. Our approach employs self-evolving agent profiles, optimized through three key metrics, guiding agents in refining their individual expertise while maintaining complementary team dynamics. $\textbf{MorphAgent}$ implements a two-phase process: a $\textbf{Profile Update}$ phase for profile optimization, followed by a $\textbf{Task Execution}$ phase where agents continuously adapt their roles based on task feedback. Our experimental results show that $\textbf{MorphAgent}$ outperforms existing frameworks in terms of task performance and adaptability to changing requirements, paving the way for more robust and versatile multi-agent collaborative systems.