Especially in open systems, development of the individual agents might not be in the hand of the system designer. In many cases, the designer of the infrastructure differ from those using those designs to create agents using them. The designers of an electronic market, e.g., do not necessarily create the agents that participate in the market und use its facilities to trade and conduct business.  Instead, the system designer has to define interfaces, data models, and interactions so that other development teams know how the agents should behave in the system, interact with other agents, and with the system as a whole. Even if the system and the individual agents are implemented by the same company different teams within the organisation can be responsible for the implementations, necessitating the same degree of separation and standardisation. If possible, the potential developers of the agent level should be part of the design process of the system level to ensure that interfaces correspond to the needs of the agent design.

In total, four different areas can be distinguished in the development of a multi-agent system. Each of these areas requires separate design artefacts, tasks, and a different focus of the stake holders:

1. Agent Architecture: the design of the individual agents, separate for each distinct type of agent.
2. Agent Organisation: the specification of organisational paradigms that will structure the agent population at runtime.
3. Agent Interaction: the definition of interfaces and protocols used by agents to exchange information, delegate control, and change organisational structure.
4. System Architecture: the relationship between the different types of agents, the supporting infrastructure, external components, and the environment.