Brain-computer interfaces (BCIs) are systems that allow us to translate in real-time the electrical activity of the brain in commands to control devices. They do not rely on muscular activity and can therefore provide communication and control for those who are severely paralyzed (locked-in) due to injury or disease. It has been shown that locked-in patients are able to achieve EEG-controlled cursor or limb movement and patients have successfully communicated by means of a BCI. Current BCIs differ in how the neural activity of the brain is recorded, how subjects (humans and animals) are trained to produce a specific EEG response, how the signals are translated into device commands, and which application is provided to the user. The present review focuses on approaches to BCIs that process the EEG on-line and provide EEG feedback or feedback of results to the user. We regard online processing and feedback cornerstones for routine application of BCIs in the field. Because training patients in their home environment is effortful and personal and financial resources are limited, only few studies on BCI long-term use for communication with paralyzed patients are available. The need for multidisciplinary research, comprising computer science, engineering, neuroscience, and psychology is now being acknowledged by the BCI community. A standard BCI platform, referred to as BCI2000, has been developed, which allows us to better combine and compare the different BCI approaches of different laboratories. As BCI laboratories now also join to unify their expertise and collaborations are funded, we consider it realistic that within few years we will be able to offer a BCI, which will be easy to operate for patients and caregivers.