Research shows that most agent-based collaborations suffer from lack of flexibility. This is due to the fact that most agent-based applications assume pre-defined knowledge of agents' capabilities and/or neglect basic cognitive and interactional requirements in multi-agent collaboration. The highlight of this paper is that it brings cognitive models (inspired from cognitive sciences and HCI) proposing architectural and knowledge-based requirements for agents to structure ontological models for cognitive profiling in order to increase cognitive awareness between themselves, which in turn promotes flexibility, reusability and predictability of agent behaviour. The ultimate aim is towards applications which advocate user-centeredness such that as little cognitive overload is incurred on humans. The semantic web is used as an action mediating space, where shared knowledge base in the form of ontological models provides affordances for improving cognitive awareness.