Taekwondo is one of the martial arts from South Korea. One of the categories in this martial arts branch is Poomsae, which contains movement techniques. This research aims to develop a system to classify Taekwondo Poomsae moves using MediaPipe. The main objective is to categorize Taekwondo accurately moves into four kick classes namely Ap Chagi, Dollyo Chagi, Yeop Chagi, and Dwi Chagi, and seven stance classes namely Arae Makki, Batangson, Eolgol Makki, Hansonal Momtong Makki, Jepipom Mokchigi, Momtong Jireugi, and Sonnal Momtong Makki. This research begins with data collection in the form of video recordings from professional Taekwondo athletes using a camera. The joints of the athlete's body are detected and localized in each frame using MediaPipe to provide sequential movement information in numerical values in the form of X, Y and Z joints. Features are then extracted from these poses using Multilevel Wavelet Packet Entropy analysis and used as input to the classification model. The method involves pose extraction, feature extraction, and training and evaluation. The developed system obtained the highest accuracy at MWPE Level 3 on the kick category dataset at 93.75%, while in the parry category, it reached 96.42%. This study advances movement and martial arts analysis by highlighting how well the suggested system works to automate the categorization of Taekwondo Poomsae motions. Future study can broaden the scope by expanding the system in Kyorugi or match categories.