A cognitive agent is an agent characterized by properties that are generally attributed to humans. Cognition is viewed here as a general mechanism of reasoning (in contrast with reactive agents) about knowledge. Such agents can perceive their environment, reason about fact or epistemic states of other agents, have a decision making process, etc. This article presents the main concepts used in cognitive agents formalizations, and speak about two particular concepts related to humans: trust and emotion. The language used for cognitive agents is here a logical language because it particularly fits well for both knowledge representation and reasoning formalization. But, even if trust and emotion can be both easily formalized by logical languages, we show that some numerical models are also well adapted.