Microservice deployment in resource-constrained edge environments remains challenging. While the architecture demonstrates adaptability, conventional methods have two critical limitations: (1) heavy reliance on manual configuration for dependency resolution and (2) computationally intensive optimization processes that hinder rapid deployment. To address these problems, we propose a large language model (LLM)-powered multi-agent microservice deployment strategy generation framework (LMAMD) that automates deployment strategy design through collaborative reasoning. Our approach leverages LLMs’ capabilities in constraint analysis, task decomposition, and distributed decision-making. Experimental validation across three real-world datasets of varying scales demonstrates that, compared with baseline methods implemented based on various concepts, the LMAMD outperforms microservice deployment strategies under different parameter conditions. This work pioneers a new direction for intelligent edge orchestration by synergizing LLM-based cognitive agents with microservice system dynamics. The code can be found at https://github.com/dongconse/LMAMD .