We present for the first time a sentence boundary detection system for identifying sentence boundaries in Vietnamese texts. The system is based on a maximum entropy model. The training procedure requires no hand-crafted rules, lexicon, or domain-specific information. Given a corpus annotated with sentence boundaries, the model learns to classify each occurrence of potential end-of-sentence punctuations as either a valid or invalid sentence boundary. Performance of the system on a Vietnamese corpus achieved a good recall ratio of about 95%. The approach has been implemented to create a software tool named vnSentDetector, a plug-in of the open source software framework vnToolkit which is intended to be a general framework integrating useful tools for processing of Vietnamese texts.