Natural language EEG lab notebook. Train pattern classifiers, analyze connectivity, compare sessions — all by talking to an AI copilot that reads your live brain signals.
Watch realistic conversations with your brain copilot
Ask about your current mental state
how's my signal?
All channels read 'good', quality 0.98 — trustworthy signal.
Train custom mental state classifiers
train a pattern called 'meditation'
Starting pattern training for 'meditation'. I'll guide you through recording segments. First, I need a baseline. Sit normally with eyes open. Ready to record 'rest' for 4 seconds? (say yes when ready)
Deep frequency-domain analysis
analyze my alpha rhythm
Spectral analysis: • Individual alpha peak (IAF): 10.2 Hz • Alpha power: 0.73 (dominant) • 1/f slope: -1.1 (typical healthy EEG) • Spectral entropy: 0.68 (moderate complexity) • Frontal alpha asymmetry: -0.05 (balanced)
Network topology and synchrony
show connectivity in the alpha band
Alpha-band connectivity (amplitude coupling): Mean correlation: 0.34 Strongest pair: C3↔C4 (r=0.72) — motor cortex synchrony Most connected: C3 (6 strong links) Least connected: Fp1 (2 weak links)
All the capabilities of a neuroscience lab, accessible through conversation
Monitor focus, relaxation, engagement with session-relative indices
Train custom classifiers with L2 + group-lasso, LORO-CV validation
IAF detection, 1/f slope, theta/beta ratio, spectral entropy
Cross-channel amplitude coupling in any frequency band
Capture labeled windows, compare with effect sizes (Cohen's d)
Eye blinks (single/double), jaw clenches, motion artifacts
Auto-generate analysis notebooks
One command. No Node.js required.
curl -sSL https://raw.githubusercontent.com/pieeg-club/PiEEG-agent/main/install.sh | bashpieeg-agent webOpen http://localhost:8000 for chat, live brain state cards, pattern training UI
pieeg-server --mock --lslpieeg-agent webDesigned for neurofeedback research and UX prototyping — not clinical use
Dedicated LSL thread, ring buffer — no sample loss even if LLM is slow
Tools request state on demand — token costs stay sane
Events, not voltage arrays — models reason about "focus_high", not floats
Swap providers, run without LLM, test with mock — debuggable, testable, maintainable
Dry-run, cooldown, audit log — AI can't spam device commands
Anthropic, OpenAI, Groq, Ollama, LM Studio — your choice, your hardware
Join researchers and makers building the future of brain-computer interfaces
Open source • CC BY-NC 4.0 License • Non-commercial use