Distributed Agent Reasoning Across Independent Systems With Strict Data Locality
The paper presents a proof-of-concept where agents in separate systems cooperate only through natural language messages, while each organisation keeps its own data local behind strict operational boundaries. The focus is on feasibility and architectural patterns rather than clinical or production evaluation.