<p>Path planning in unknown, multi-floor buildings presents significant challenges for both autonomous systems and human navigation. Navigating such environments requires reaching a target location without prior knowledge of the building’s layout or obstacles. In particular, firefighters face these challenges during emergencies, where low-visibility conditions and restricted movement complicate navigation. This paper introduces BUG-FIRE, a 3D path planning algorithm designed to assist firefighter navigation in single-tower buildings without prior environmental knowledge. BUG-FIRE combines BUG2, an algorithm widely used in autonomous robotic systems, and adapts it for first responder navigation, using it for intra-floor movement. For inter-floor transitions, we integrate an emergency exit exploration algorithm that ensures goal convergence while complying with building code standards, such as mandatory stairwell connectivity and emergency signage. We validate BUG-FIRE through 250 Monte Carlo simulations in a virtual building model, demonstrating a 100% success rate within the tested simulation scenarios for reaching randomly selected targets across multiple floors. The results show that agents with a 30-meter vision radius reduce path length by 28.2 compared to those with only 0.5-meter visibility, highlighting the importance of environmental perception. To bridge the gap to real-world applications, we implement a YOLOv8-based vision system on the Magic Leap 2 AR headset, enabling real-time emergency signage recognition and dynamic AR guidance. Experimental results indicate that users can successfully navigate 3D paths utilizing AR guidance, with trajectory lengths closely matching simulated benchmarks.</p>