This paper presents the operational semantics of an agent-oriented programming language called CLAIM. CLAIM allows to design multi-agent systems that support both stationary and mobile agents, endowed with cognitive capabilities and able to communicate. The semantics takes into account agents' reasoning, mobility, communication and concurrence, and contains a set of reduction rules between coherent states of a program. The operational semantics is a first step towards the formal verification of agent-based programs written in CLAIM. The reduction rules are illustrated on a case study.