ken-you-reflect

A simple, universal MCP tool that challenges any claim when you want to verify it.

Purpose

This MCP provides one tool: reflect - which asks a verification question when you suspect an agent might be making assumptions, lying, or being overconfident about something.

Universal: Works for coding, business, creative work, research, or any domain where agents make claims.

Installation

Prerequisites

• Node.js 16 or higher
• npm or yarn

Setup

1. Clone this repository

2. Install dependencies:

   npm install

3. Build the project:

   npm run build

Adding to Claude Code

Add this server to Claude Code by running:

claude mcp add ken-you-reflect -s user -- node /absolute/path/to/dist/index.js

How It Works

Pure Manual: You manually trigger it when you want to challenge a claim. The tool always returns the same universal verification question.

// When you hear something suspicious:
await reflect({
  claim: "This approach is proven to work",
  context: "Building new feature",
});

// Always returns the same universal question:
// 🚨 VERIFICATION REQUIRED: What evidence supports this claim?

Universal Examples

Coding

await reflect({
  claim: "According to the React docs, we should use useEffect here",
  context: "Adding component lifecycle",
});
// Always returns: "What evidence supports this claim?"

Business

await reflect({
  claim: "Market research shows this strategy will increase revenue by 20%",
  context: "Quarterly planning",
});
// Always returns: "What evidence supports this claim?"

Creative Work

await reflect({
  claim: "This narrative structure is proven to engage readers",
  context: "Writing a novel",
});
// Always returns: "What evidence supports this claim?"

Research/Analysis

await reflect({
  claim: "Studies demonstrate this approach reduces user churn",
  context: "Product optimization",
});
// Always returns: "What evidence supports this claim?"

When to Use It

Use ken-you-reflect whenever you want to challenge a claim:

• Coding: Agent references documentation, APIs, or best practices
• Business: Claims about market research, user behavior, or strategies
• Creative: Assertions about what works in writing, design, or art
• Research: Claims about studies, data, or proven methods
• Personal: Any advice or recommendations that sound too confident

The Universal Question

The tool always asks the same powerful question that works across all domains:

"What evidence supports this claim?"

This single question is effective because it:

• Forces verification of any assertion
• Works in coding, business, creative, and research contexts
• Is simple and predictable
• Requires no complex logic or maintenance

API Reference

reflect

Challenge any claim with a verification question.

Parameters:

• claim (string, required): The claim to challenge and verify
• context (string, optional): Optional context about the situation

Returns: The universal verification question "What evidence supports this claim?"

Philosophy

Manual Control: You decide when to challenge claims, not some algorithm trying to guess.

Universal Application: Works for any domain - coding, business, creative, research, personal advice.

Universal Question: Always asks the same powerful question that works everywhere.

Maximum Simplicity: No complex logic, no maintenance burden, completely predictable.

The goal is simple: Give you a tool to challenge suspicious claims across any domain with one powerful question.

Development

npm install
npm run build
npm test
npm run lint:check

License

MIT