Application domains were central to Agentcities because a real testbed demanded real use cases. It was not enough to prove that platforms could connect; the network had to show that agents could solve problems people actually cared about. Several working groups organized themselves around specific domains, producing both research insights and practical demonstrations. This application focus is what distinguished Agentcities from pure theory — it was a live infrastructure for demonstrating agents tackling genuine, consequential problems rather than toy examples.
Healthcare Applications
The Healthcare Working Group recognized that patient care is inherently a multi-party coordination problem. Physicians, nurses, laboratories, pharmacies, insurers, and administrative systems each hold only partial information and carry distinct responsibilities, yet they must act in concert. Agent-based approaches offered autonomous coordination across these parties without forcing every system into a single monolithic database. The group's focus areas included clinical data integration across heterogeneous systems, patient monitoring with autonomous alert agents that watch for deteriorating conditions, resource scheduling for scarce assets such as operating theatres, imaging equipment, and specialist time, and clinical decision support. The First International Workshop on Health Care Applications of Intelligent Agents, held alongside iD3 in Barcelona on February 7–8, 2003, drew researchers from across Europe and demonstrated working prototypes that grounded these ideas in realistic clinical scenarios.
Emergency Rescue
The Rescue Working Group addressed one of the most demanding coordination problems imaginable: emergency response under extreme time pressure, with distributed resources, damaged infrastructure, and multiple organizations that may never have worked together before. Autonomous agents could coordinate search teams, direct resource allocation, and maintain shared situational awareness without relying on a centralized command structure that might itself be destroyed or cut off. The RoboCup Rescue simulation league ran in parallel within the wider research community, providing a common benchmark for these ideas. The key challenges the group confronted were coordination under deep uncertainty, graceful degradation when communication channels fail, and establishing trust between agents that represent different organizations with different authorities and operating procedures.
Business Process Integration
The Business Process Integration Working Group tackled enterprise-scale coordination: how could agents mediate between organizations with different systems, data formats, business rules, and levels of trust? Traditional Electronic Data Interchange (EDI) required bilateral agreements and bespoke, hand-built integrations for every pairing of partners. Agent-based approaches envisioned something far more dynamic — semantic mediation in which an agent representing one organization's procurement system could discover and interact with an agent representing another organization's inventory system without any pre-existing bilateral integration in place. This anticipated much of what would later be called the API economy and service-oriented architecture, but with a richer semantic grounding that let systems negotiate meaning rather than merely exchange pre-agreed message formats.
Security in Open Agent Networks
Security was a cross-cutting concern, relevant not to one application domain but to the entire open network. The Security Working Group addressed three intertwined questions: trust, or how an agent decides whether to accept a request from an unknown counterpart; authentication, or how an agent verifies another's identity; and authorization, or what a given agent is actually permitted to do. Trust reasoning — an agent's ability to evaluate the trustworthiness of potential interaction partners based on reputation, recommendations, and context — was a distinctive research contribution of this group. Contemporary AI agent systems face nearly identical security questions about delegation, authority, and trust in open environments, making this early work strikingly relevant to today's autonomous-agent deployments.
The questions these working groups raised about coordination, trust, and security in open, multi-party systems are once again at the forefront of computer science as autonomous AI agents proliferate. The National Institute of Standards and Technology publishes frameworks for AI risk management and security that build on decades of distributed systems security research — the same intellectual lineage that the Agentcities Security and Rescue working groups helped to advance.