MCP Systems Setup Repo

This repository demonstrates how to set up and use the Model Composition Protocol (MCP) with TinyLlama in VS Code.

Overview

The Model Composition Protocol (MCP) enables AI systems to expose capabilities to other systems in a standardized way. This project shows how to:

• Set up a local MCP server using TinyLlama
• Create and use MCP clients
• Accelerate MCP development using LLMs
• Integrate MCP with VS Code

Prerequisites

• Python 3.13+
• devbox for development environment management
• Git

Getting Started

1. Clone this repository
2. Initialize the development environment:

   devbox shell

3. Install dependencies using uv (automatically handled by devbox)

Running the Server

Start the basic MCP server with:

python simple_server.py

Or try the Rick knowledge base server:

python rick_server.py

This will initialize the TinyLlama model and expose it through an MCP server running on localhost.

Testing with a Client

Connect to the basic server using:

python simple_client.py

Or connect to the Rick knowledge base server:

python rick_client.py

These demonstrate basic interactions with the MCP server, including tool calls and resource utilization.

Simple Llama Chat

For a direct interface with the TinyLlama model without MCP:

python simple_llama_chat.py

MCP Inspector

For debugging and exploring the MCP server capabilities, use the MCP Inspector:

uv run mcp dev simple_server.py

Or for the Rick server:

uv run mcp dev rick_server.py

Integration with Claude

To install your MCP server capabilities in Claude:

uv run mcp install simple_server.py

Guides and Tutorials

• Building MCP with LLMs - Detailed instructions on using Claude and other frontier LLMs to accelerate your MCP development
• MCP Tutorial - Step-by-step tutorial for getting started with MCP

Project Structure

• simple_server.py - Basic MCP server implementation with TinyLlama
• simple_client.py - Basic MCP client for testing
• rick_server.py - MCP server with Rick and Morty knowledge base
• rick_client.py - Client for the Rick knowledge base server
• Ricks_KB.txt - Knowledge base text file for the Rick server
• simple_llama_chat.py - Direct interface to TinyLlama without MCP
• building_mcp_with_llms.md - Guide for MCP development with LLMs
• mcp-tutorial.md - Step-by-step MCP tutorial
• devbox.json - Development environment configuration
• devbox.lock - Locked dependencies for devbox
• pyproject.toml - Python project configuration
• uv.lock - Locked dependencies for uv package manager
• models/ - Directory for downloaded model files (gitignored)
• test/ - Test files for the project
• __pycache__/ - Python bytecode cache (gitignored)

Resources

• MCP Documentation
• TinyLlama on Hugging Face