Brain-computer interfaces (BCIs) for treatment of spinal cord injury (SCI) can be used to reanimate paralyzed limbs. Most approaches stimulate the peripheral nerves or muscles, which has limitations. To overcome these challenges, here we show that a BCI can control spinal stimulation and improve forelimb function in rats with cervical SCI. We decoded forelimb movement intention via multichannel local field potentials in sensorimotor cortex using a canonical correlation analysis, which is both less computationally complex and more stable over time than spike-based algorithms. We then used this decoded signal to modulate epidural stimulation and restore forelimb movement. Finally, we implemented the efficient closed-loop algorithm in a miniaturized on-board computing platform. This Brain-Computer-Spinal Interface (BCSI) approach utilized recording and stimulation approaches already used in separate clinical trials, so it may readily translate to human subjects as a potential solution to upper extremity paralysis.