This paper describes the use of coherence constraints as a means to regulate agent interaction. Coherence constraints describe relationships between the content of utterances, and the context. They can be used for example to express that an answer must refer back in a meaningful way to the question that it answers. We also discuss several possible ways in which the enforcement of coherence constraints can be implemented in a multi-agent system. Finally we describe a possible implementation in the 3APL platform, which shows the feasibility of this form of interaction regulation.