The identification of cacao (Theobroma cacao) is essential for quality assurance, optimized postharvest practices, and traceability within the cacao value chain. This study focuses on the development of software solution that performs real-time classification of cacao pod varieties based on captured images. A dataset of cacao pods from different farms in Silang, Cavite was collected and annotated for model training, incorporating variations in lighting, pod maturity. The system utilizes a lightweight convolutional neural network (CNN) architecture specifically MobileNetV2 optimized for on-device inference via TensorFlow Lite, enabling offline classification without constant internet access. It was evaluated by 30 respondents composed of local cacao farmers, college instructors and I.T. professionals using the ISO 25010 Software Quality Metrics and obtained descriptive rating of “Very Good”. The application offers a scalable, accessible for cacao variety identification, contributing to improved quality control, value-based marketing, and the broader digital transformation of the Philippine cacao industry.