This chapter discusses the necessary representations for a distributed flexible agent-based approach. Some agents will control internal computation processes, for example, an agent providing constraint computations based on kernel methods, and will use only deep representations. Other agents will have the task of producing surface representations to interface with humans, as for example, do personal assistants. Irrespective of whether agents are used to support the design and manufacturing, different kinds of information must be represented. General theoretical models have been introduced mostly in the context of design. Such models described the design process as a set of successive steps and tried to give a mathematical axiomatization to ground the design activity. In some of the modeling approaches, knowledge is handled explicitly. A multi-agent system for concurrent engineering must be composed of a number of different agents: agents providing specific application services; control agents; and agents interfacing with users.