Spam texts have become a common problem with mobile communication’s exponential expansion, generating trouble and possible security concerns. This paper offers a quick SMS spam detection tool using machine learning methods. The system combines sophisticated text preprocessing, TF-IDF-based feature extraction, and an optimized XGBoost classifier to separate spam from valid messages properly. Model performance is improved, and data imbalance is handled using hyperparameter tuning and weighted loss functions. A dynamic message classification tool lets users predict interactively and in real time. The strength of the suggested system is shown by performance evaluation using accuracy, F1-score, recall, precision, and the lift metric. The findings show that the model has great classification accuracy, therefore qualifying it as a consistent spam detection tool for mobile communication.