Multiple Choice Questions (MCQs) are used to assess students’ knowledge, but developing them takes a lot of time and effort. This paper presents AutoTest, an Agentic AI framework powered by GPT-3.5-turbo that uses the capabilities of GPT-3.5-turbo to automatically generate UPSC-style MCQs. In addition, the system uses Retrieval-Augmented Generation (RAG) to perform web searches to verify all data to make certain the questions appear cussing and relevant. Because of this, the degree of manual work involved has diminished while the whole process is more scalable and remains high quality via dynamic content generation and online validation. This framework is perfectly suitable for educators and educational institutions, as it can save them a huge amount of time and resources. With this technology, educators would be able to spend more of their energy on teaching than on administrative tasks such as MCQ generation. MCQs generated by this technique can, in fact, be very quickly updated and validated using the most recent real-time data, keeping the MCQs relevant and current with regard to the latest changes for a better student assessment.