Stroke is the leading cause of disability in the United States, and yet no definitive interventions can drive the nervous system beyond its measurable but often limited spontaneous recovery. Treatment to limit injury progression and enhance repair after stroke or other types of central nervous system injury is complicated by the heterogeneous nature of cell death and wound healing mechanisms and the multiple barriers to functional recovery. The heterogeneity of injury and repair mechanisms requires interventions that are broad and multi-modal, but also intrinsically safe. We describe a process to identify such interventions by screening multiple individual targets in the historically separate realms of neuroprotection, repair, and regeneration against a library of FDA-approved compounds with known safety. We have identified nearly 10 compounds that are able to activate simultaneously protective and reparative genes. These compounds have a theoretical therapeutic window that spans from evolving injury (minutes to hours) to stable injury (days to months to years). It is our hypothesis that these compounds will be most efficacious when paired with training. The notion is that drugs will alter the propensity of the nervous system toward recovery, whereas specific training will engage the needed instructive cues to achieve this goal. Indeed, robotic training can provide a level of motor learning that seems to enhance the salutary effects of training. In a system that depends heavily on cells that do not easily replenish themselves, cellular therapies could also ultimately be an important part of the cocktail. We conclude that combinations of interventions will be needed to surmount the multiple barriers to recovery in stroke and other types of brain and spinal cord injury recovery.