In previous talks we (Dave, Jarred & myself) have outlined a small communication language for expressing multi-agent dialogues. This language is intended to be a successor to the restrictive finite-state based models currently used in Electronic Institutions. Until now, our treatment of this language has been largely informal, and lacking in an implementation. In my talk I will outline a partial formal specification of this language and demonstrate my implementation of this specification in Java. I will show several agent dialogues in action and discuss the problems associated with the provision of a full formal specification.