This paper describes a practical, low-cost research design for a non-invasive brain-machine interface (BMI) that enables a human operator to command a small quadcopter (a “bird drone”) using external EEG signals only (no invasive implants). The aim is a proof-of-concept that demonstrates safe, limited-scope mind-in-the-loop control suitable for accessibility demonstrations, rehabilitation motivation, and research into embodied agency. The design uses robust, well-studied EEG paradigms (SSVEP and motor-imagery) for discrete command selection, consumer / research-grade EEG headsets (OpenBCI / Emotiv), open-source flight stacks (PX4 / ArduPilot), and lightweight telemetry middleware to translate decoded intent into high-level flight modes (e.g., takeoff/hover/land/left/right/forward/back). Emphasis is placed on layered safety: geofencing, low altitude ceilings, engageable fail-silent autopilot modes, manual RC override, and ethical review. The paper provides system architecture, recommended components, data-flow diagrams, signal processing/ML options, validation protocol, limitations, and an implementation checklist. This is intended as a research & education blueprint — not a consumer product — and recommends responsible institutional review and local regulatory compliance prior to any flight tests.