Online learning continues to expand due to globalization and the COVID-19 pandemic. However, maintaining student engagement in this new normal has become increasingly difficult. Conventional techniques, such as self-reports and manual observations, often fall short of capturing the subtle behaviors that indicate attentiveness. This emphasizes the necessity for sophisticated tools to assess engagement effectively. The proposed system introduces an innovative approach to monitoring student attention in online learning environments by integrating computer vision techniques with a Gradient Boosting classifier (GBC). It conducts a multimodal analysis of behavioral cues captured through a standard webcam, such as facial expressions, hand movements, mobile phone usage, and head poses, to enable a comprehensive and accurate evaluation of student engagement. With thorough validation on a dataset of 6,000 records, the GBC model outperformed traditional approaches and other machine learning algorithms by attaining an accuracy of 99.13%. Through the utilization of Explainable AI (XAI) tools such as LIME and SHAP, we increased the transparency and interpretability of our model. This allows educators to gain a better understanding of the elements that influence student engagement, hence promoting trust among all stakeholders involved. The system's focus on resource efficiency and scalability makes it adaptable to diverse educational settings without extensive infrastructure. The user-friendly web interface facilitates real-time monitoring, seamlessly integrating with popular e-learning platforms and providing detailed, anonymized reports. This enables instructors to make data-driven interventions to enhance teaching strategies and offers actionable insights to improve learning outcomes. Non-identifiable data collection meets ethical requirements while maintaining privacy and producing insightful engagement metrics.