Semantic Prompt MCP

Semantic Prompt MCP

The Core Thinking Engine for SuperClaude Framework - An MCP server that helps Claude think systematically step-by-step

🎯 What is this?

Semantic Prompt MCP is a tool that helps Claude solve complex problems by breaking them down into 3-4 systematic thinking steps.Just like humans solve problems by "understanding first → choosing a method → executing", this makes Claude follow the same process.

Core Concepts

• 🧠 Step-by-Step Thinking: Break complex problems into 3-4 manageable steps
• 📚 Document Caching: Once read, documents are cached and referenced (performance optimization)
• 🎨 Profile System: Different configuration files for different purposes

⚡ Quick Start

1. Run without Installation

npx semantic-prompt-mcp

2. Add to Claude Code

Add to .mcp.json in your project root:

{
  "mcpServers": {
    "semantic-prompt": {
      "command": "npx",
      "args": ["-y", "semantic-prompt-mcp@latest"],
      "env": {
        "CHAIN_OF_THOUGHT_CONFIG": "superclaude.json"
      }
    }
  }
}

🎭 Three Modes (Profiles)

1️⃣ default.json - Basic Mode

npx semantic-prompt-mcp  # or
npx semantic-prompt-mcp default.json

• Purpose: General problem solving
• Features: Flexible thinking process, simple 3-step structure
• Best for: General tasks without special framework requirements

2️⃣ superclaude.json - SuperClaude Mode ⭐

npx semantic-prompt-mcp superclaude.json

• Purpose: Use with SuperClaude Framework
• Features:
  • 90% command selection enforcement (systematic execution)
  • Document duplicate read prevention (caching system)
  • 21 dedicated commands support
  • Quality Gates validation system
• Best for: Required when using SuperClaude Framework

3️⃣ supergemini.json - SuperGemini Mode

npx semantic-prompt-mcp supergemini.json

• Purpose: Use with SuperGemini Framework
• Features:
  • 4-step structure (Analysis → TOML Command → Agent Selection → Execution)
  • Commands defined in TOML files
  • Multi-Agent system support
• Best for: When using SuperGemini Framework

🔄 How It Works

SuperClaude Mode Example (3 Steps)

User: "Analyze security issues in this code"

Step 1️⃣ - Intent Analysis
Claude: "User wants security analysis. Related files are..."

Step 2️⃣ - Command Selection (90% enforced)
Claude: "Selecting 'analyze' command and reading analyze.md document"
System: Provides analyze.md content → Cache saved ✅

Step 3️⃣ - Execution Strategy
Claude: "Following document instructions to execute security analysis..."

🚀 Core Feature: Document Caching System

Documents are never read twice!

First Request:
Claude: "Selecting 'analyze' command"
System: Reads analyze.md → Cache saved ✅

Second Request:
Claude: "Selecting 'analyze' command"
System: "Already read. Refer to system-reminder" ⚡

This significantly reduces token usage and speeds up execution!

🎨 Creating Your Own Custom Profile

Step 1: Create your custom JSON file

Create a new file my-custom.json in any folder (e.g., your project root):

{
  "tool": {
    "name": "my_thinking",
    "description": "My custom thinking process..."
  },
  "config": {
    "availableCommands": ["analyze", "build", "test"],
    "commandPath": "./my-commands/",
    "commandPreference": 0.8
  }
}

Step 2: Use it in Claude Code

Update your .mcp.json:

{
  "mcpServers": {
    "semantic-prompt": {
      "command": "npx",
      "args": ["-y", "semantic-prompt-mcp@latest"],
      "env": {
        "CHAIN_OF_THOUGHT_CONFIG": "./my-custom.json"  // ← Just change this!
      }
    }
  }
}

That's it! Just change the filename in CHAIN_OF_THOUGHT_CONFIG:

Built-in profiles (no path needed):

• "superclaude.json" - SuperClaude Framework
• "supergemini.json" - SuperGemini Framework
• "default.json" - Basic mode

Your custom profiles (need path):

• "./my-custom.json" - File in your project root
• "./config/my-profile.json" - File in a subfolder
• "/absolute/path/to/profile.json" - Absolute path

Why the difference? Built-in profiles are packaged with npm, your files are in your project!

📁 Project Structure

semantic-prompt-mcp/
├── prompts/
│   ├── default.json      # Basic profile
│   ├── superclaude.json  # SuperClaude specific
│   └── supergemini.json  # SuperGemini specific
├── src/
│   └── index.ts          # Main server code
├── LICENSE              # MIT License
└── README.md            # This document

🤝 For Developers

Local Development Setup

git clone https://github.com/hyunjae-labs/semantic-prompt-mcp.git
cd semantic-prompt-mcp
npm install
npm run build
npm link  # For local testing

🔧 Troubleshooting

"Document already read" message appears

This is normal! Documents are cached for performance optimization.

Too many console logs

export DISABLE_THOUGHT_LOGGING=true

Cannot find specific command

Check your command path:

export CHAIN_OF_THOUGHT_COMMAND_PATH=/correct/path/to/commands/

📜 License & Attribution

This project is based on sequential-thinking MCP server.

License

MIT License - Free to use, modify, and distribute

Copyright Notice

• Original Work: Copyright (c) Model Context Protocol Contributors (sequential-thinking)
• This Work: Copyright (c) 2025 Hyunjae Lim ([email protected])

Major Changes from Original

• Extended 3-step structure to adaptive 3-4 step structure
• Added SuperClaude/SuperGemini Framework specific profiles
• Implemented document caching system
• Added meta-cognitive attention mechanisms
• Implemented multi-profile system

🔗 Related Links

• Model Context Protocol
• Semantic Prompt MCP Repository
• SuperClaude Framework
• SuperGemini Framework
• Original sequential-thinking

🚀 Version History

v1.3.0 (Current)

• Added SuperClaude/SuperGemini profiles
• Implemented document caching system
• Added meta-cognitive attention mechanisms

v1.0.0

• Initial release (based on sequential-thinking)

💡 Need Help?

• Open an issue: GitHub Issues
• Refer to SuperClaude Framework or SuperGemini Framework
• Contact: Hyunjae Lim ([email protected])