A large body of evidence from animal studies indicates that motor intention can be decoded via multiple single-unit recordings or from local field potentials (LFPs) recorded not only in primary motor cortex, but also in premotor or parietal areas. In humans, reports of invasive data acquisition for the purpose of BCI developments are less numerous and signal selection for optimal control still remains poorly investigated. Here we report on our recent implementation of a real-time analysis platform for the investigation of ongoing oscillations in human intracerebral recordings and review various results illustrating its utility for the development of novel brain-computer and brain-robot interfaces. Our findings show that the insight gained both from off-line experiments and from online functional exploration can be used to guide future selection of the sites and frequency bands to be used in a translation algorithm such as the one needed for a BCI-driven cursor control. Overall, the findings reported with our online spectral analysis platforms (Brain TV and Brain Ball) indicate the feasibility of online functional exploration via intracranial recordings in humans and outline the direct benefits of this approach for the improvement of invasive BCI strategies in humans. In particular, our findings suggest that current BCI performance may be improved by using signals recorded from various systems previously unexplored in the context of BCI research such as the oscillatory activity recorded in the oculomotor networks as well as higher cognitive processes including working memory, attention, and mental calculation networks. Finally, we discuss current limitations of the methodology and outline future paths for innovative BCI research.