Nowadays, in Latin America, a huge amount of people are in a motor disability condition. This phenomenon generates difficulties to execute daily tasks, such as the feeding process. To mitigate the daily difficulties, assistance devices are needed. This paper describes the evaluation of a brain-computer interface (BCI) for meal assistance, based on the sensorimotor rhythm (SMR), characteristic of the movement intention. The electroencephalographic (EEG) signals were acquired and processed to extract features in the frequency and time domain. These features train a classification model that separates the movement intention from any other cerebral activity. The study was made with ten healthy people who were subjected to a test that corresponds to feed themselves ten times. The results obtained show that average time to activate the meal assistance device is less than 10 s, furthermore, the accuracy of the tests performed was 81.6%, i.e. there is a good differentiation between a movement intention from another activity. Finally, it was concluded that the purposed meal assistance device achieves the goal of allowing an autonomous feeding and leaves as a precedent an alternative to implementing assistance devices through a BCI.