In this paper we address the verification of communication between agents participating in multi-agent web service systems. Our approach is founded on the application of model-checking techniques to protocols which express interactions between a group of agents in the form of a dialogue. We outline a web service architecture which supports the construction of multi-agent systems using web services technology. We then define a lightweight protocol language which can express a wide range of inter-agent dialogues, and we use the SPIN model checker to verify properties of this language. Our initial results show this approach has a satisfactory success rate in the detection of failures.