When presented with identical task conditions across multiple trials, the brain produces variable patterns of neural activity. This neural variability stems from many sources, one of which is fluctuations in internal states (e.g., arousal, attention, and motivation). Neural variability may reflect changes in these internal cognitive factors and can influence our perceptual and decision-making abilities. For instance, when tasked with shooting consecutive free throws on a basketball court, we may experience changes in our arousal, and our minds might wander. Such trial-to-trial variability can detrimentally impact our task performance by limiting information encoding in our brains. Also, deficits in regulating neural variability have been linked to neuropsychiatric disorders. Being able to reduce neural variability thus has implications in enhancing and restoring the brain’s cognitive capabilities. However, the extent to which neural variability is under volitional control is unclear. We designed a prefrontal cortex (PFC) brain-computer interface (BCI) to assess whether a macaque could use moment-to-moment visual feedback to stabilize its neural activity. We challenged the subjects to use the BCI to keep their neural activity as close as possible to a baseline neural state observed at the beginning of each session. We discovered that subjects successfully used the moment-to-moment visual feedback to produce neural activity similar to baseline activity. Throughout the session, we observed that the neural activity gradually moved away from the initial baseline state. While non-BCI trials exhibited these slow fluctuations in neural activity, we found that BCI trials suppressed these slow changes. Overall, our results demonstrate that subjects could suppress neural variability using our novel neurofeedback paradigm. Furthermore, these findings can inform the development of clinical BCIs that treat cognitive disorders.