PV Bhat Vibe Check MCP Server

🚀 Vibe Check

Vibe Check is a tool designed to address the disorderly behavior issues caused by "mental inertia" during the operation of large language model (LLM) agents. By integrating three core functions: vibe_check, vibe_distill, and vibe_learn, this project provides a mechanism that enables LLM agents to actively identify and correct deviations in their reasoning processes during runtime.

✨ Features

1. vibe_check

• Purpose: Enforce stage-consistent interruptions when the agent starts explaining basic blockchain concepts (such as blocks, linked lists, etc.) to develop a to-do application.
• Trigger Conditions: When the agent deals with concepts or logic unrelated to the current task.
• Implementation Mechanism: Embed checkpoints in each reasoning stage (planning/implementation/review) to ensure the effectiveness of interruptions.

2. vibe_distill

• Purpose: Enforce compression and simplification when the agent's plan contains excessive nested points (e.g., detailed to the level of a technical specification document).
• Trigger Conditions: When the agent generates overly complex solutions or plans.
• Implementation Mechanism: Convert complex logic into a concise chain of key points as the basis for recalibration.

3. vibe_learn

• Purpose: Record and analyze error patterns after the agent gets stuck in excessive complexity to optimize subsequent interruption strategies.
• Trigger Conditions: After the user manually guides the agent back from complexity.
• Implementation Mechanism: Establish a feedback loop for storing and utilizing historical error data.

📦 Installation

Automatic Installation via Smithery

To automatically install the vibe-check-mcp-server for Claude Desktop via Smithery:

npx -y @smithery/cli install @PV-Bhat/vibe-check-mcp-server --client claude

Manual Installation (Recommended)

# Clone the repository
git clone https://github.com/PV-Bhat/vibe-check-mcp-server.git
cd vibe-check-mcp-server

# Install dependencies
npm install

# Build the project
npm run build

# Start the server
npm run start

📚 Documentation

Integration with Claude

Add the following content to your claude_desktop_config.json file:

"vibe-check": {
  "command": "node",
  "args": [
    "/path/to/vibe-check-mcp/build/index.js"
  ],
  "env": {
    "GEMINI_API_KEY": "YOUR_GEMINI_API_KEY"
  }
}

Environment Configuration

Create a .env file in the project root directory:

GEMINI_API_KEY=your_gemini_api_key_here

Agent Prompt Guide

To effectively implement pattern interruptions, include the following instructions in the system prompt:

As an autonomous agent, you need to:
1. Treat vibe_check as a crucial pattern interruption mechanism.
2. Always include the complete user request in each call.
3. Specify the specific reasoning stage (planning/implementation/review).
4. Use context-relevant interruption strategies.

Usage Scenarios

Example 1: Developing a To-Do Application

{
  "name": "todo-app",
  "description": "A simple todo application built with Node.js.",
  "planning": {
    "features": ["CRUD operations", "User authentication"],
    "technologies": ["Node.js", "Express.js", "MongoDB"]
  }
}

Example 2: Teaching Blockchain Basics

{
  "name": "blockchain-lesson",
  "description": "Teaching blockchain basics to new developers.",
  "planning": {
    "topics": ["Block structure", "Blockchain validation"],
    "examples": ["Bitcoin transaction flow", "Ethereum smart contracts"]
  }
}

API Reference

1. vibe_check

• Purpose: Embed interruption checkpoints in the planning, implementation, or review stages.
• Input Parameters:
  • stage: The current reasoning stage ("planning" | "implementing" | "reviewing").
  • context: Context information (e.g., the current task description, executed steps, etc.).
• Return Value: Interruption strategy or continue signal.

2. vibe_distill

• Purpose: Simplify complex reasoning processes into a chain of key points.
• Input Parameters:
  • complexity_level: Complexity index (1 - 5).
  • content: Context information to be simplified.
• Return Value: A list of simplified key points.

3. vibe_learn

• Purpose: Analyze and record error patterns in the reasoning process.
• Input Parameters:
  • error_type: Error type ("off-topic" | "overly complex" | "logical error").
  • context: Context information when the error occurred.
• Return Value: Feedback signal or optimization suggestions.

Documentation References

Installation Guide

1. Automatic Installation via Smithery
2. Manual Installation (Recommended)

Integration Guide

1. Integration with Claude
2. Environment Configuration

API Reference

1. vibe_check
2. vibe_distill
3. vibe_learn

Project Structure

.
├── README.md          # Project documentation
├── package.json      # Project dependencies and scripts
├── src/
│   ├── index.js     # Main program entry
│   └── utils/
│       ├── check.js  # Implementation of vibe_check
│       ├── distill.js # Implementation of vibe_distill
│       └── learn.js   # Implementation of vibe_learn
└── .env.example      # Example of environment variables

💻 Usage Examples

Example 1: Developing a To-Do Application

{
  "name": "todo-app",
  "description": "A simple todo application built with Node.js.",
  "planning": {
    "features": ["CRUD operations", "User authentication"],
    "technologies": ["Node.js", "Express.js", "MongoDB"]
  }
}

Example 2: Teaching Blockchain Basics

{
  "name": "blockchain-lesson",
  "description": "Teaching blockchain basics to new developers.",
  "planning": {
    "topics": ["Block structure", "Blockchain validation"],
    "examples": ["Bitcoin transaction flow", "Ethereum smart contracts"]
  }
}

⚠️ Important Note

• Environment Dependencies: Ensure that Node.js and npm are installed.
• Permission Settings: Ensure that you have the correct file read and write permissions.
• Error Handling: It is recommended to add error capturing and logging mechanisms in a production environment.

🤝 Contribution Guidelines

Any contributions are welcome! Please refer to the CONTRIBUTING.md file to learn how to participate.

📄 License

MIT License

📞 Contact

If you have any questions, please contact author's email or visit the project homepage.