We present an AI architecture used for modeling computational rational cognitive agents--- agents that can reason, and act in an environment based on their representations of beliefs, acts, and intentions. A unified representational formalism is employed for representation of propositions (beliefs), goals (desires), actions, and plans. The representational formalism is described using a conceptual object-oriented hierarchy. An object-oriented design facilitates a uniform method (message) protocol that is naturally amenable to concurrency. The reasoning and acting behavior of the modeled agent is defined by an actor-like [Agha 1986, Agha and Hewitt 1987] message passing scheme. Semantically speaking, messages can be viewed as meta-level propositional attitudes (corresponding to 'belief' and 'intention') that are sent to/from propositions and acts in parallel. The message passing model, called a Rational Engine (as opposed to an inference engine), implements a BDI-architecture.