Brain-computer interface (BCI) spellers offer a non-muscular communication method for a wide range of users, from healthy individuals to those with motor impairments. Although numerous studies have been conducted on the topic, there is a lack of localization for Vietnamese. As such, this paper proposes an end-to-end BCI speller system for Vietnamese citizens using the Motor Imagery (MI) paradigm. The system employs a machine learning (ML) model to continuously classify MI tasks and converts these classifications into commands for controlling a virtual Vietnamese keyboard.