PockeTA is a pedagogical agent using AI, accessible on and off campus. It can be used in a variety of situations, such as on PCs for studying and on mobile devices during field trips. The pedagogical agent, supported by a multi-agent system, is deeply integrated with a graph-based knowledge base, K-Cube, which is managed by teachers. The system collects student questions, organizes them into learning-analytics metrics, and visualizes them for instructors to help them monitor students’ progress and learning experiences. In this paper, we outline the framework of our system and share some promising early results. Positive feedback from early adopters suggests PockeTA may help address TA scalability; rigorous controlled studies are required to evaluate effects on learning outcomes and TA workload. These challenges are particularly evident in many regions worldwide, where universities often struggle to improve the TA-to-student ratio due to resource constraints and growing student populations.