Cerebral palsy (CP) is the leading cause of motor disability in children worldwide, yet access to continuous, high-quality physiotherapy remains limited, especially in low- and middle-income settings. This study presents the design and technical validation of a mobile physiotherapy assistant that leverages Google MediaPipe’s BlazePose for real-time human-pose estimation to support home-based rehabilitation for children with CP. The system adopts a client–server architecture with a Flutter front-end, Flask back-end, and MySQL database to deliver low-latency video processing and quantitative feedback. Rule-based algorithms transform 3-D body-landmark data into clinically relevant metrics such as joint angles, postural stability, and transitional movement parameters, enabling six progressive therapy stages inspired by the Gross Motor Function Measure. Internal testing demonstrated accurate kinematic measurements (mean absolute error <5°) and stable real-time performance (average 11–13 fps) across all therapy modules. These findings confirm the feasibility of using commodity devices for asynchronous, family-centered telerehabilitation.