TheoremExplainAgent is a multimodal AI system that generates long-form Manim animation videos to explain mathematical theorems, demonstrating deep LLM understanding while uncovering reasoning gaps that text-only explanations miss. The system leverages retrieval-augmented generation (RAG) and LLM agent techniques to produce and evaluate video-based explanations. The codebase includes generation code, evaluation metrics, and a dataset of theorem-video pairs for benchmarking.