Pose Anything is a graph-based general pose estimation method designed to make keypoint localization applicable to any object category using a single model with minimal support image annotations. The method leverages the geometric relationships between keypoints through a novel graph transformer decoder, improving keypoint localization accuracy. Pose Anything has demonstrated outstanding performance on the MP-100 benchmark, surpassing previous state-of-the-art techniques, achieving significant improvements in 1-shot and 5-shot settings. Compared to previous CAPE methods, its end-to-end training exhibits scalability and efficiency.