Theory of Knowledge,

Augmentable Intelligence

Tokai fuses EEG brain-computer interface data with agentic AI to help neurodivergent minds focus, plan, and perform — continuously and empirically.

NTU Master’s research projectOpen source · Apache 2.0Neuroethics & privacy-firstStar us on GitHub

Built for how your brain actually works

Traditional productivity tools ignore your biology. Tokai reads it.

Nine-Metric Neural Dashboard

Track Focus Index, Bio Energy, Mental Fatigue, Working Memory Load, Theta/Beta Ratio, and four more cognitive signals — live. Works from a 30-second self-report, simulated data, research EEG datasets, or a real BCI headset.

Agentic Task Planner

TokAgent (powered by Claude) reads all nine neural metrics alongside your tasks, journal, and medication log — then takes direct action in your workspace: creating tasks, starting timers, logging entries, all from a natural-language command.

Privacy-First Architecture

Neural data never leaves your browser. All cognitive metrics are held in session memory only — never written to a server. Behavioral data (tasks, notes, medications) is isolated per-user via row-level security at the database level.

How Tokai works

Three steps from cognitive state awareness to a smarter workday.

Choose your data mode

Start with a 30-second self-report check-in — no hardware required. Switch to simulated mode for demos, dataset replay for research, or plug in an EEG headset for live neural streaming.

Your cognitive state comes alive

Nine metrics update in real time — Focus Index, Bio Energy, Mental Fatigue, Working Memory Load, and more — giving you an empirical picture of where your brain is right now.

TokAgent adapts your workflow

The AI planner recommends the right task for your current cognitive state, executes commands directly in your workspace, and keeps a transparent audit trail of every action it takes.

What you’ll see

Your Focus Index, live.

Tokai computes a continuous Focus Index from your chosen data source — self-report, simulated, EEG dataset replay, or live BCI — giving you an empirical cognitive baseline without requiring any hardware to get started.

Detect deep focus vs. cognitive fatigue in real time
Track Alpha/Beta wave ratios and neural noise
See hyperfocus risk and estimated focus window duration

Focus Index

Live

84/ 100

Optimal focus window · 38 min remaining

Bio Energy

72%

Neural Noise

Low

α/β Ratio

0.41

Mental Fatigue

WM Load

Hyperfocus

12%

TokAgent

High focus detected. Recommend starting your most cognitively demanding task now.

Inside Tokai v0.2.0

Six integrated tools, all driven by your neural data.

Neural Dashboard

Nine live cognitive metrics — Focus Index, Sleep Quality, Bio Energy, Mental Fatigue, Working Memory Load, Neural Noise, T/B Ratio, Focus Window, and Hyperfocus Risk — each tagged with the data source driving it. BCI-only metrics dim automatically in self-report mode.

TokAgent

A Claude-powered AI assistant that reads all nine neural metrics alongside your task list, journal, and medication log — then takes direct action via ten tool-calling actions. Uses Claude Sonnet 4.6 for planning, Haiku 4.5 for fast inline recommendations. Bring your own API key; it never touches Tokai's servers.

TokDo

A focus-aware task manager where each task carries a cognitive demand score matched against your live Focus Index. Set time estimates, deadlines, and emoji tags — TokAgent uses all of this to surface the right task at the right moment.

TokNote

An ADHD-friendly journal with mood tagging and focus logging. Capture how you feel throughout the day and let TokInsights surface correlations between your journal entries and cognitive patterns over time.

TokMed

Medication and supplement tracking with timestamp-anchored markers on your focus charts, so you can see exactly how each dose correlates with your cognitive state throughout the day.

TokInsights

New

Automatic, device-local analysis of your productivity patterns — when you focus best, mood correlations with task completion, and weekly cognitive trends. All analysis runs in your browser; no data leaves your device.