Getting Started

How to integrate this AI constitution for BCI projects into your development workflow

⚡ Instant Setup (Recommended)

The fastest way to get started - 3 seconds:

# Install as dev dependency
npm install --save-dev @the-governor-hq/constitution-bci

Then run setup:

npx governor-install-bci

This configures:

✅ .cursorrules - Cursor AI safety rules for BCI data
✅ .vscode/settings.json - GitHub Copilot instructions
✅ .mcp-config.json - MCP server for Claude/ChatGPT
✅ package.json - Adds ai:context script

Your AI coding assistant is now safety-aware for BCI/neurotechnology data development.

Other domains: See constitution-wearables (opens in a new tab), constitution-therapy (opens in a new tab), or constitution-core (opens in a new tab)

Verify Installation

Check what was configured:

# Should see .cursorrules in your project root
ls -la .cursorrules
 
# Should see MCP config
ls -la .mcp-config.json
 
# Should see ai:context script in package.json
npm run ai:context

For MCP-Compatible AI (Claude Desktop, etc.)

Start the MCP server to provide context to external AI assistants:

npm run ai:context

Or add to Claude Desktop config:

macOS: ~/Library/Application Support/Claude/config.json
Windows: %APPDATA%\Claude\config.json

{
  "mcpServers": {
    "governor-hq-bci": {
      "command": "node",
      "args": ["./node_modules/@the-governor-hq/constitution-bci/dist/mcp-server.js"]
    }
  }
}

Manual Setup (Alternative)

If you prefer manual configuration or can't use npm:

Step 1: Add to Your Project

Clone or link this repository alongside your BCI data project:

# Option A: Git submodule
cd your-project/
git submodule add https://github.com/the-governor-hq/constitution.git docs/governor-hq
git submodule update --init --recursive
 
# Option B: Clone separately
cd ~/projects/
git clone https://github.com/the-governor-hq/constitution.git governor-hq-docs

Step 2: Configure AI Assistant Context

When using AI coding assistants, include this repository in your context. Different tools have different methods:

GitHub Copilot / VS Code

Add to your workspace or project configuration:

// .vscode/settings.json
{
  "github.copilot.chat.codeGeneration.instructions": [
    {
      "file": "packages/bci/README.md"
    }
  ]
}

Cursor

Add to project's .cursorrules file:

# BCI Data Safety Constitution
Follow the Governor HQ Constitutional Framework for BCI data projects.
Never generate code that:
- Makes neurological or psychiatric diagnoses
- Claims to read thoughts or emotions from brain signals
- Recommends medical treatments or interventions
- Uses brain data without explicit informed consent
- Makes cognitive ability claims from EEG patterns

Claude / ChatGPT

In your conversation context, include:

"I'm working on a consumer BCI/neurotechnology product that processes brain data (EEG, neurofeedback, etc.). I need to follow the Governor HQ Constitutional Framework for BCI safety. Key constraints: no neurological diagnoses, no thought/emotion reading claims, no cognitive ability assessment, no medical treatment suggestions. Reference documentation at: [link to docs]"

What Gets Installed

Configuration Files

The package creates/updates these files in your project root:

`.cursorrules`

Safety instructions for Cursor AI assistant specific to BCI data projects.

`.vscode/settings.json`

GitHub Copilot instructions for BCI safety constraints.

`.mcp-config.json`

MCP server configuration for Claude Desktop and other MCP-compatible tools.

Scripts Added to package.json

{
  "scripts": {
    "ai:context": "node ./node_modules/@the-governor-hq/constitution-bci/dist/mcp-server.js"
  }
}

Using the Framework

During Development

When implementing BCI features, ask your AI assistant:

"Does this comply with the BCI constitutional framework?"
"Check this notification text against BCI language rules"
"Validate this neurofeedback feature against hard rules"
"Is this brain pattern interpretation allowed?"

Code Review Checklist

Before merging BCI-related code:

No neurological diagnoses (ADHD, epilepsy, dementia, etc.)
No cognitive ability claims (intelligence, learning disabilities)
No emotion/thought reading assertions
No medical treatment recommendations
User-facing text uses observational language, not diagnostic
Personal baseline established before generating insights
Privacy and consent explicitly handled
Brain data described as patterns, not mind reading

BCI-Specific Development Workflow

Phase 1: Data Collection (0-30 days)

Silent learning period. System observes brain patterns but provides no recommendations.

if (user.bciBaselineStatus !== 'STABLE') {
  return null; // No recommendations during baseline learning
}

Phase 2: Pattern Recognition (30-90 days)

Baseline established. System can now notice meaningful deviations.

const isSignificantDeviation = 
  user.currentAlphaPower < user.personalBaseline.alphaPower * 0.8;
 
if (isSignificantDeviation && user.bciBaselineStatus === 'STABLE') {
  return generateNeurofeedbackSuggestion(user);
}

Phase 3: Personalized Guidance (90+ days)

Mature understanding. System has rich context about individual patterns.

Testing Your Integration

Quick Test

# Run the built-in tests
npm test
 
# Should validate:
# - MCP server starts correctly
# - Configuration files exist
# - AI context is accessible

Manual Validation

Ask your AI assistant: "What are the BCI constitutional framework rules?"
Expected response: Should reference hard rules, no diagnoses, pattern recognition only, etc.
If it doesn't know: Check MCP server is running or configuration files are loaded

Continuous Integration

Add to your CI pipeline:

# Example GitHub Actions workflow
- name: Validate BCI Safety Compliance
  run: |
    npm run test
    # Add custom linting for forbidden terms
    ! grep -r "diagnose\|ADHD\|epilepsy\|read.*thoughts" src/

Troubleshooting

AI Assistant Doesn't Seem Aware of Rules

Check configuration files exist:
```
ls .cursorrules .mcp-config.json
```
Restart your IDE/editor
For MCP tools: Verify the server is running
```
npm run ai:context
```

Rules Seem Too Restrictive

That's intentional. BCI data is:

Highly sensitive (neural activity)
Easy to misinterpret (patterns ≠ thoughts)
Legally risky (medical claims)
Ethically complex (privacy, consent)

If a feature feels blocked, ask: "Is this truly consumer wellness, or is it medical?"

Examples of Safe BCI Features

✅ Focus training game

Neurofeedback when user enters focused state
Personal baseline required
No cognitive ability claims

✅ Meditation assistant

Detects relaxation patterns in alpha/theta waves
Guides user toward calm states
No mental health diagnosis

✅ Sleep stage tracker

Estimates sleep stages from EEG
Shows personal sleep architecture
No sleep disorder diagnosis

❌ ADHD attention trainer → Claims medical diagnosis
❌ Lie detector → Claims thought reading
❌ IQ estimator → Claims cognitive ability measurement

Next Steps

Read the AI Agent Guide for detailed coding patterns
Review Hard Rules to understand absolute boundaries
Check What We Don't Do for scope clarity
Consult domain-specific agents: Focus, Neurofeedback

Support

Introduction AI Agent Guide