Multi-robot systems (MRS) require efficient path planning to navigate complex environments while avoiding collisions. While traditional heuristic algorithms and modern learning-based methods have made significant strides, they often face challenges with computational scalability, adaptability, or semantic reasoning in dynamic environments. This paper presents a novel framework that leverages a Large Language Model (LLM) for waypoint generation in a 2D corridor environment with two mobile robots. The environment consists of axis-aligned corridors, junctions, and obstacles, all encoded into a structured textual prompt. The LLM generates coordinate-based waypoints for both robots simultaneously, and a closed-loop feedback mechanism validates these waypoints against spacing, collision, and junction constraints. Invalid waypoints trigger iterative refinement until all constraints are satisfied. Experimental results demonstrate that the proposed approach achieves high waypoint generation success rates while maintaining path feasibility in complex scenarios.