This thesis presents the design and implementation of a lightweight AI-powered penetration testing tool that demonstrates how automation and artificial intelligence can enhance cybersecurity assessment. The system integrates traditional scanning methods, such as network and web application analysis, with a rule-based AI engine that evaluates vulnerabilities, assigns risk scores, and generates remediation recommendations. Built in Python using a modular architecture, each component operates independently to ensure flexibility, extensibility, and clarity in function. The project emphasizes explainable AI rather than complex machine learning, making it suitable for educational and experimental use. Comprehensive testing confirmed the tool’s reliability, accuracy, and efficiency, highlighting its value as both a practical learning framework and a foundation for future advancements in automated security testing.