This dataset comprises continuous 2D trajectory decoding from attempted arm movements recorded via 60-channel EEG from 10 participants (9 able-bodied, 1 with spinal cord injury) across 3 sessions. Participants performed motor imagery tasks including target-tracking (snakerun) and self-paced shape tracing (freerun) with visual feedback. The preprocessed data, acquired at 200 Hz and downsampled to 20 Hz, includes temporal features and source localization estimates suitable for regression-based brain-computer interface applications in neuroprosthetics and upper limb restoration.