learnforge-mcp by MarcoPWx - MCP Server

GenericMCP

🤖 A generic Model Context Protocol (MCP) server template - Build your own MCP integrations with local LLMs

🌟 Features

This generic MCP server provides a foundation for building custom AI integrations, using 100% local LLMs via Ollama for complete privacy:

🔍 Search Capabilities: Implement custom search across your data sources
🌐 Content Processing: Process and generate content with local LLMs
📝 Dynamic Generation: Create custom content using LLMs (Qwen, Llama, etc.)
🎯 Custom Workflows: Build your own AI-powered workflows
📊 Data Management: Track and manage your custom data
🚀 Extensible Framework: Easy to extend with your own tools

🏗️ Architecture

┌─────────────┐     MCP Protocol    ┌───────────────────┐
│Claude/AI    │◄───────────────────►│GenericMCP         │
│Assistant    │                     │Server             │
└─────────────┘                     └───────┬───────────┘
                                            │
                ┌───────────────────────────┼───────────────────────────┐
                │                           │                           │
        ┌───────▼────────┐       ┌─────────▼────────┐       ┌──────────▼──────┐
        │Your Data       │       │Ollama Local LLMs │       │Custom Tools     │
        │Sources         │       │Qwen/Llama/Mistral│       │& Integrations   │
        └────────────────┘       └──────────────────┘       └─────────────────┘

📦 Installation

Prerequisites

Node.js 18+
Ollama installed and running
Your custom data sources (optional)
Additional tools as needed for your use case

Quick Start

Clone this repository:

git clone https://github.com/yourusername/GenericMCP.git
cd GenericMCP

Install dependencies:

npm install

Install Ollama and pull a model:

# Install Ollama (macOS)
brew install ollama

# Start Ollama service
ollama serve

# Pull Qwen 2.5 (recommended) or any other model
ollama pull qwen2.5:7b

Test the server:

npm start

🔧 Configuration

Environment Variables

export DATA_PATH="/path/to/your/data"
export TOOLS_PATH="/path/to/your/tools"
export OLLAMA_HOST="http://localhost:11434"
export OLLAMA_MODEL="qwen2.5:7b"

Claude Desktop Integration

Add to ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "generic-mcp": {
      "command": "node",
      "args": ["/path/to/GenericMCP/index.js"],
      "env": {
        "DATA_PATH": "/path/to/your/data",
        "TOOLS_PATH": "/path/to/your/tools",
        "OLLAMA_MODEL": "qwen2.5:7b"
      }
    }
  }
}

🛠️ Available Tools

Tool	Description	Example
`search`	Search your data sources	"Find relevant documents"
`process_content`	Process content with LLMs	"Analyze this text"
`generate`	Generate content using AI	"Create a summary"
`custom_workflow`	Execute custom workflows	"Run my automation"
`manage_data`	Manage your data	"Update records"
`extend_tools`	Add new capabilities	"Add custom function"

🔐 Privacy & Security

✅ 100% Local LLMs - No cloud API calls
✅ Ollama Integration - Run on your hardware
✅ No Telemetry - Zero tracking
✅ Offline Capable - Works without internet
✅ Open Source - Fully auditable

🚀 Supported Models

Model	Size	Best For
qwen2.5:7b	4.7GB	General purpose
llama3.2:3b	2.0GB	Lightweight
mistral:7b	4.1GB	Code generation
deepseek-coder	4.1GB	Programming

📄 License

MIT License - See file

Built with ❤️ for privacy-conscious AI learners