Although many systems for palliative communication based on non-invasive BCIs have been developped during the last few years, very few projects aim at leaving the research labs and hospitals for helping patients at home. Jon Wolpaw's team at the Wadworth Center1 has developped a portable BCI that has now been used for more than one year on a daily basis by 5 people suffering from ALS. This experiment shows that highly handicaped people greatly benefit from such BCIs that they tend to use during long periods -- between 5 an 8 hours a day -- for communicating with their loved ones, for surfing the web, or reading and writing emails. We also aim at being able to leave the lab with this now mature technology for screening handicaped people at home. This would allow checking easily if a patient can use efficiently a BCI without requiring him to come to the hospital or to a specialized laboratory. From the hardware point of view, this home-screening requires a BCI setup that can be used in any situation: portable, fully autonomous and battery powered. From the software point of view, the machine learning techniques that adapt the BCI to the individual must provide a "good" result within a few seconds rather than an "optimal" result after several minutes or hours of processing.