Public Template for Docker MCP LLM GUI Application

This repository contains both demo applications from my recent Docker MCP YouTube Tutorial — combining Docker MCP Catalog, remote & local MCP servers, and a full-stack LLM web app using Docker Model Runner.

This is a further development of my Full Stack LLM Project from: https://github.com/MariyaSha/simple_AI_assistant

🎥 Video Tutorial

To setup and run this application step by step, please watch my YouTube tutorial below. It will explain all the app processes, and the logic behind them.

This video demonstrates how to:

• connect AI tools like Claude and Cursor to real services using Docker MCP.
• embed MCP tools into your own Python application.
• and build a production-ready full-stack AI app that calls MCP tools via a unified Docker gateway.

If you watched the video, this is the codebase used in the workflow.

🧠 What This App Does

• Uses the Docker MCP Gateway to unify access to both local and remote MCP servers.
• Calls real MCP tools (e.g., Stripe, DuckDuckGo, Hugging Face) from a Python app.
• Shows how Docker manages authentication, endpoints, and tool discovery.
• Builds a full-stack GUI AI app powered by Docker Model Runner + MCP.

This app serves as a starter template for building professional AI assistants with real tools.

🚀 Getting Started

Prerequisites

• Docker Desktop (with MCP Toolkit & Catalog)
• Python 3.12+
• (Optional) Stripe account for remote Stripe MCP

📦 Setup Instructions - MCP Demo App (MCP calls only)

1. Clone the repository

git clone https://github.com/MariyaSha/Docker_MCPGUIApp.git
cd Docker_MCPGUIApp/mcp_demo_app

2. Edit catalog.yaml and add your Stripe Secret Key:

Authorization: "Bearer <Stripe secret key goes here>"

3. Build and Run Application

docker compose up --build

📦 Setup Instructions - Complete App (Full Stack LLM Web Assistent)

1. Clone the repository

git clone https://github.com/MariyaSha/Docker_MCPGUIApp.git
cd Docker_MCPGUIApp/complete_app

2. Configure environment variables

Rename the example file:

mv example.env .env

Edit .env and add your secrets:

STRIPE_SECRET_KEY="<YOUR SECRET KEY FROM STRIPE GOES HERE>"

3. Install a model in Docker Desktop

• Open Docker Desktop
• Go to Models
• Install a lightweight model (e.g. Gemma3)
• Update the model name in .env if you go for a different model

4. Build and run the app

docker compose up --build

Open your browser at:

http://localhost:8501

🧪 What You’ll See

• Calls to local MCP tools (DuckDuckGo) if you prompt for "search X on the web"
• Calls to remote MCP tools (Hugging Face) if you prompt for "search X on HuggingFace"
• Unified access via Docker MCP Gateway
• A full-stack AI web interface - the perfect template to develop your own LLM apps.

📚 Resources

• ⭐ Docker MCP Catalog: https://dockr.ly/4pzVKGd

💡 Notes

• Remote MCP servers authenticate via browser-based OAuth.
• Local MCP servers require API keys configured in Docker Desktop.
• Docker MCP Gateway still requires API Key authentication - may change in the future as Docker MCP is currently still in Beta.
• Applications only talk to Docker MCP, not individual MCP servers.

❤️ Contributing

Feel free to fork, extend, and reuse this project.

🧑‍💻 License

MIT License