Advances in large language models are enabling increasingly capable multi-agentic systems, raising important questions about how robots should represent, coordinate, and communicate the behaviours of multiple underlying agents. As these systems become more complex and opaque, new challenges arise for autonomy, transparency, cognitive load, and users' mental models when robots serve as embodied interfaces to distributed agentic processes.

The MAgicS-HRI workshop brings together researchers to explore these challenges through design activities, discussions, and paper presentations. By examining real-world scenarios and their associated tensions, the workshop aims to outline practical design considerations and future research directions for creating intelligible and trustworthy multi-agentic robots that can meaningfully support end-users.