Detecting gas leaks is essential to ensure safety in residential, industrial, and commercial environments. This paper introduces an intelligent gas monitoring system using the MQ-6 semiconductor sensor with an ESP32 microcontroller, integrated with a Long Short-Term Memory (LSTM)-based deep learning model. Time-series sensor data is continuously collected and preprocessed using normalization and sliding-window segmentation methods. The trained LSTM model captures temporal variations in the data to provide early and precise predictions of potential leaks. Once a leak is identified, automated safety measures such as solenoid valve closure, power disconnection, and audible alarms are triggered to minimize risks. Furthermore, sensor readings and prediction results are uploaded in real time to a Firebase cloud platform, ensuring remote supervision, data storage, and instant notifications. Experimental evaluations show that the proposed system outperforms conventional threshold-based methods, offering higher sensitivity and reliability. The integration of artificial intelligence with Internet of Things (IoT) technology highlights a scalable and proactive solution for advanced gas safety management.