Prompt Gen Mcp

PromptGen MCP - Local Server

Transform simple questions into comprehensive, context-aware prompts using advanced prompt engineering techniques fetched from PromptGen API and your local code context.

🚀 Quick Start

1. Get Your PromptGen API Key

1. Visit: https://promptgenmcp-production.up.railway.app
2. Sign up and generate your API key
3. Copy your API key starting with pg_sk_

2. Get Other Required API Keys

• GROQ API Key: Get from https://console.groq.com/
• Tavily API Key: Get from https://tavily.com/

3. Install Dependencies

# Navigate to the directory
cd prompt-gen-mcp

# Install dependencies
pip install mcp sentence-transformers httpx groq

4. Configure Cursor

Copy the configuration to your Cursor MCP settings file (~/.cursor/mcp_servers.json):

{
  "mcpServers": {
    "prompt-gen": {
      "command": "python",
      "args": ["/full/path/to/prompt-gen-mcp/src/prompt_gen_mcp/server.py"],
      "env": {
        "PROMPTGEN_API_KEY": "pg_sk_your_key_here",
        "GROQ_API_KEY": "gsk_your_groq_key_here", 
        "TAVILY_API_KEY": "tvly_your_tavily_key_here"
      }
    }
  }
}

Important: Replace /full/path/to/prompt-gen-mcp/ with the actual full path to this directory.

5. Restart Cursor

Completely restart Cursor (not just reload) for the MCP server to be recognized.

6. Start Using

1. Open any project in Cursor
2. Press Cmd+Shift+P (Mac) or Ctrl+Shift+P (Windows/Linux)
3. Type: "MCP: enhance_prompt"
4. Ask: "How do I optimize this React component?"

💡 What You Get

🔄 Transformation Example

Before	After
"How do I optimize this React component?"	Comprehensive Enhanced Prompt

Enhanced Output:

# 🚀 Enhanced Prompt for: How do I optimize this React component?

## 🔍 Question Analysis
- **Types**: optimization, implementation  
- **Complexity**: medium
- **Domain**: frontend
- **Intent**: optimize performance

## 🎯 Selected Prompt Engineering Techniques

### 1. 📋 Plan and Solve (Score: 0.847)
**Description**: Break down complex problems into structured planning and systematic solving phases

**Application**: Plan: Identify optimization targets. Solve: Apply specific React patterns.

### 2. 📚 Few-Shot Learning (Score: 0.823)
**Description**: Provide examples to guide the response

**Application**: Here are optimization examples: useMemo, React.memo, useCallback...

## 📁 Relevant Code Context

### File 1: `src/components/UserProfile.tsx` (Relevance: 0.892)
```typescript
const UserProfile = ({ userId }) => {
  const [user, setUser] = useState(null);
  useEffect(() => {
    fetchUser(userId).then(setUser);
  }, [userId]);
  return <div>{user?.name}</div>;
};

✨ Enhanced Analysis Request

[Structured instructions for comprehensive response with context-aware suggestions...]

📊 Feature Comparison

Feature	Basic Prompt	Enhanced Prompt
Context Awareness	❌ None	✅ Full codebase context
Technique Selection	❌ Manual	✅ AI-powered optimal selection
Privacy	⚠️ Varies	✅ Code stays local
Structured Output	❌ Basic	✅ Comprehensive analysis
Code Examples	❌ Generic	✅ From your actual project
Relevance Scoring	❌ None	✅ Semantic similarity matching

🔒 Architecture & Privacy

• 🏠 Local Processing: All code scanning happens on your machine
• 🌐 API Integration: Only technique selection uses PromptGen API
• 🔢 No Code Shared: Your code never leaves your machine
• 🎯 Smart Selection: AI chooses optimal techniques from hosted database
• 📁 Context Aware: Automatically includes relevant code from your project

⚙️ How It Works

Architecture Overview

flowchart TD
    A[User Question] --> B[Local MCP Server]
    B --> C[Code Context Scanner]
    B --> D[PromptGen API]
    C --> E[Local Code Analysis]
    D --> F[Technique Selection]
    E --> G[Enhanced Prompt]
    F --> G
    G --> H[User Receives Enhanced Prompt]
    
    style A fill:#e1f5fe
    style B fill:#f3e5f5
    style C fill:#e8f5e8
    style D fill:#fff3e0
    style G fill:#fce4ec
    style H fill:#e0f2f1

Interaction Flow

sequenceDiagram
    participant U as User
    participant M as MCP Server
    participant C as Code Scanner
    participant P as PromptGen API
    participant Q as Qdrant DB
    
    U->>M: Ask question
    M->>C: Scan local workspace
    M->>P: Request techniques for question
    P->>Q: Query vectorized techniques
    Q-->>P: Return relevant techniques
    C-->>M: Local code context
    P-->>M: Optimal techniques
    M->>M: Combine context + techniques
    M->>U: Enhanced prompt with context
    
    Note over C,M: Code never leaves your machine
    Note over P,Q: Only question text sent to API

Privacy-First Flow:

1. Local MCP scans your code privately on your machine
2. PromptGen API provides optimal techniques from hosted Qdrant database
3. Local processing combines everything into enhanced prompt
4. No code leaves your machine - only question text sent to get techniques

🎯 Use Cases

• Code Optimization: Get structured analysis with relevant code context
• Architecture Decisions: Receive comprehensive comparisons with examples
• Debugging Help: Get systematic debugging approaches with code snippets
• Learning: Understand concepts with examples from your actual codebase

🔧 Troubleshooting

MCP Server Not Working:

# Test the server directly
python src/prompt_gen_mcp/server.py

# Check dependencies
pip install mcp sentence-transformers httpx groq

API Key Issues:

• Ensure PROMPTGEN_API_KEY starts with pg_sk_
• Get valid GROQ key from https://console.groq.com/
• Get valid Tavily key from https://tavily.com/

Cursor Not Finding MCP:

• Ensure complete Cursor restart (not just reload)
• Check ~/.cursor/mcp_servers.json syntax
• Verify file paths are absolute
• Look for MCP status in Cursor's bottom status bar

No Code Context:

• Ensure you're in a project directory
• Check file permissions
• Verify supported file types (.py, .js, .ts, .tsx, .jsx, .md, .txt)

📊 Features

• ✅ PromptGen API Integration (hosted technique database)
• ✅ Local Code Context Scanning (100% private)
• ✅ Privacy-Preserving Architecture
• ✅ Intelligent Technique Selection
• ✅ LLM-Powered Question Analysis
• ✅ True MCP Architecture (self-contained)
• ✅ Fallback Support (works offline)

🏗️ Architecture Benefits

Proper MCP Design:

• ✅ Single self-contained server process
• ✅ Cursor manages server lifecycle automatically
• ✅ No manual background service management
• ✅ Clean startup/shutdown with Cursor

API Integration:

• ✅ Uses PromptGen API for techniques (no local llms.txt)
• ✅ Hosted Qdrant database with vectorized techniques
• ✅ User authentication via API keys
• ✅ Usage tracking and billing support

🤝 Contributing

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

📄 License

MIT License - see LICENSE file for details.