AbstractCorpus-based speech act study has become a heated topic in recent pragmatic research. However, most research is based on traditional text corpora, and few multimodal corpora have been utilized so far. To fill this gap, the article discusses the construction of a multimodal corpus of speech acts in Chinese situated discourse. In this toolkit description, scheme design, working definition, annotation evaluation, data representation, and possible usage of such a multimodal corpus are introduced. The multimodal corpus approach to pragmatic studies, with the new analytic methods and descriptions of the data, leads to a new inquiry of multimodal pragmatics, which can provide pragmatics with the fresh perspective and new practices.