Inspecting underfloor heating systems such as Ondol poses unique challenges due to their concealed nature, spatial complexity, and fragmented communication across stakeholders. Traditional inspection workflows often lack intelligent, context-aware planning, leading to inefficiencies and oversight during design and maintenance phases. To address this gap, this study introduces a novel framework that integrates a large language model (LLM)-based AI agent with Building Information Modeling (BIM) environments through the Model Context Protocol (MCP). The agent autonomously interprets spatial and parametric data within BIM, simulates inspection scenarios, and supports context-sensitive decision-making. The framework utilizes CSV-based data storage to manage inspection metadata, enabling seamless information retrieval and updating. Inspection steps and contextual outputs are visualized directly within the Revit environment, offering intuitive feedback for inspectors and engineers. Validation was conducted through simulation and scenario-based evaluation, demonstrating the agent’s potential in automating inspection planning and enhancing communication across disciplines. This work lays the foundation for AI-augmented digital workflows in traditional HVAC system monitoring.