Programming learning is of paramount importance for students in computer-related disciplines. However, conventional learning approaches often suffer from inefficiencies and delayed feedback. This paper designs and implements an intelligent programming learning support system based on a Proactive Insight Agent (PIA). Beyond providing fundamental features such as online code editing, execution, and testing, the system’s core innovation lies in introducing a PIA-driven cognitive error correction mechanism and proactive dialogue intervention. The PIA can perceive learners’ behavioral patterns in real-time, analyze their learning states, and, upon detecting potential learning impasses, proactively initiate context-aware dialogues by deeply integrating with Large Language Models (LLMs). These dialogues guide learners through cognitive reflection and error rectification. This paper elaborates on the system’s design philosophy, architecture, and the implementation of its core functionalities. Furthermore, it demonstrates the system’s potential in assisting learners to overcome programming obstacles and enhance learning efficiency through illustrative examples.