Diabetes management requires continuous monitoring of blood glucose levels to prevent severe health complications. Conventional glucose monitoring methods rely on invasive finger-prick techniques, which are often painful, inconvenient, and discourage frequent testing. To address these limitations, this paper presents GlucoBreath, a non-invasive glucose monitoring system that estimates blood glucose levels through breath analysis. The system detects acetone concentration in exhaled breath using MQ-series gas sensors, which correlates with glucose metabolism. A NodeMCU-based microcontroller processes the sensor data and transmits it to a cloud platform via IoT communication. Machine learning algorithms are employed to analyze the collected data and predict glucose levels with improved accuracy. The predicted values are displayed through a userfriendly mobile or web interface, enabling real-time monitoring and alert notifications for abnormal conditions. The proposed system is portable, cost-effective, and eliminates the need for blood sampling, thereby enhancing user comfort and compliance. Experimental results demonstrate that the system provides reliable glucose estimation with minimal latency, making it a promising solution for non-invasive diabetes management in both home and clinical environments..