Electrocorticographic (ECoG) signals have emerged as a prominent neural interface signal modality due to their high bandwidth and availability in human subjects. We present a system for wireless recording of micro-ECoG activity in a primate performing reach-to-grasp movements. The system is comprised of a head-mounted interface, off-the-shelf receiver module, and custom software written in Labview for real-time data monitoring and storage. The head-mounted interface is composed of a custom-designed VLSI neural recording front, a commercially available FSK transmitter module, a digital interface, and a battery. The system offers a fixed gain of 40 dB, programmable bandwidth settings in the 0.1 Hz to 8.2 kHz range, digital gain of 1-16, and ADC resolution of 8-12 bits. The interface consumes 6.7 mA of current from a 3.7 V battery and transmits digitized data at 1 Mbps rate. The system offers less than 0.25% dropped packets at 3m non-line-of-sight distance. We then used the wirelessly recorded ECoG signal from the dorsal premotor cortex region to decode the movement state of the animal. The ECoG spectral features could decode the movement state, achieving close to 70% accuracy as early as 100 ms prior to actual movement onset. Our system offers a new avenue for future ECoG-based brain-machine interface systems.