AI Master Control Program (MCP) Server

The AI MCP Server enables AI models, including locally hosted models with Ollama and Claude Desktop, to interact with your computer system. It acts as a bridge that allows AI models to:

• Execute system commands
• Create, read, update, and delete files
• Control other programs
• Communicate with each other

Architecture

The system consists of:

1. MCP Server: Central server that processes requests from AI models
2. Client Library: Enables easy integration with AI models
3. Model Connectors: Interfaces with various AI model backends (Ollama, Claude Desktop, etc.)
4. Task Execution Engine: Performs system operations and program control

Installation

Prerequisites

• Python 3.8+
• Ollama (optional, for local model hosting)
• Claude Desktop (recommended default model)

Automated Installation

For quick and easy installation, use the provided installation script:

# Clone the repository
git clone https://github.com/GrizzFuOnYou/master_mcp_server.git
cd master_mcp_server

# Run the installation script
python install.py

The installation script will:

1. Verify Python version compatibility
2. Install all dependencies
3. Create a directory structure
4. Configure environment variables
5. Create platform-specific startup scripts
6. Set up Claude Desktop as the default AI model

Manual Setup

If you prefer manual installation:

1. Clone the repository:

   git clone https://github.com/GrizzFuOnYou/master_mcp_server.git
   cd master_mcp_server
   

2. Install dependencies:

   pip install -r requirements.txt
   

3. Configure environment variables:

   cp .env.example .env
   # Edit .env with your preferred settings
   

Usage

Starting the Server

Using Startup Script (Recommended)

After installation:

• Windows: Run start_mcp_server.bat
• Linux/Mac: Run ./start_mcp_server.sh

Manual Start

Run the MCP server:

python startup.py

By default, the server will listen on 0.0.0.0:8000.

Connecting AI Models

Claude Desktop (Default)

Claude Desktop is configured as the default model. To use it:

1. Make sure Claude Desktop is running on your system
2. The server will automatically attempt to connect on startup
3. Claude Desktop should be available at the default location: http://localhost:5000/api

If you need to manually connect:

from mcp_client import MCPClient

# Initialize client
client = MCPClient("http://localhost:8000", "your-secret-api-key")

# Connect to Claude Desktop
result = client.connect_model("claude-desktop", "claude", {"api_url": "http://localhost:5000/api"})
print(f"Connection result: {result}")

Claude Desktop Connection JSON

If you need to manually configure Claude Desktop integration, use the following JSON configuration:

{
  "model_id": "claude-desktop",
  "model_type": "claude",
  "config": {
    "api_url": "http://localhost:5000/api",
    "temperature": 0.7,
    "max_tokens": 1000
  }
}

Ollama Models

To connect to an Ollama model:

from mcp_client import MCPClient

# Initialize client
client = MCPClient("http://localhost:8000", "your-secret-api-key")

# Connect to an Ollama model
result = client.connect_model("llama2", "ollama", {"host": "http://localhost:11434"})
print(f"Connection result: {result}")

Executing System Operations

Once connected, AI models can perform various system operations:

# Execute a command
result = client.execute_system_command("claude-desktop", "echo", ["Hello, World!"])

# Write a file
result = client.write_file("claude-desktop", "test.txt", "This is a test file created by Claude!")

# Read a file
result = client.read_file("claude-desktop", "test.txt")

# Start a program
result = client.start_program("claude-desktop", "notepad.exe")

# Stop a program
result = client.stop_program("claude-desktop", pid)

# Query the AI model
result = client.query_model("claude-desktop", "claude-desktop", "What is the capital of France?")

API Reference

Server Endpoints

Endpoint	Method	Description
/connect_model	POST	Connect to an AI model
/disconnect_model/{model_id}	POST	Disconnect from an AI model
/list_models	GET	List all connected models
/execute_task	POST	Execute a task requested by an AI model
/task_status/{task_id}	GET	Get the status of a task

Client Methods

Method	Description
connect_model(model_id, model_type, config)	Connect to an AI model
disconnect_model(model_id)	Disconnect from an AI model
list_models()	List all connected models
execute_system_command(model_id, command, args, working_dir, timeout)	Execute a system command
execute_file_operation(model_id, operation, path, content)	Execute a file operation
control_program(model_id, action, program_path, args, pid)	Control a program
query_model(model_id, target_model, prompt)	Query an AI model

Model Configuration

Claude Desktop Configuration

To connect to Claude Desktop, use the following configuration:

{
  "api_url": "http://localhost:5000/api",
  "temperature": 0.7,
  "max_tokens": 1000
}

Ollama Configuration

To connect to an Ollama model, use the following configuration:

{
  "host": "http://localhost:11434"
}

Security Considerations

IMPORTANT: This server grants AI models significant access to your system. Use with caution.

Security measures implemented:

• API key authentication
• Logging of all operations
• Configurable permissions (coming soon)
• Rate limiting (coming soon)

Troubleshooting

Claude Desktop Connection Issues

If you encounter issues connecting to Claude Desktop:

1. Ensure Claude Desktop is running
2. Verify the API URL (default: http://localhost:5000/api)
3. Check the logs for specific error messages
4. Restart Claude Desktop and try again

Ollama Connection Issues

If you encounter issues connecting to Ollama:

1. Ensure Ollama is running (ollama serve)
2. Verify the model exists (ollama list)
3. Check the API URL (default: http://localhost:11434)
4. Try pulling the model again (ollama pull modelname)

Extension Points

The MCP server can be extended to support:

• Additional AI model backends
• More sophisticated program control
• GUI interaction capabilities
• Web browsing capabilities
• Network operation capabilities

License

MIT

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.