We present an architecture for a rational, reactive agent and describe its implementation. The paper addresses issues raised by the interaction of the rational and reactive behaviour of the agent, and its updating mechanism. We relate it with the work of others.