Azure Container Apps - AI & MCP Playground

Name: Azure Container Apps - AI & MCP Playground
Author: manekinekko

@manekinekko

About Azure Container Apps - AI & MCP Playground

This project showcases how to use the MCP protocol with Azure OpenAI. It provides a simple example to interact with OpenAI's API seamlessly via an MCP server and client.

#azure #mcp #openai #agents #llm

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "azure-container-apps-ai-mcp": {
      "command": "docker",
      "args": [
        "compose",
        "up",
        "-d",
        "--build"
      ]
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is Azure Container Apps - AI & MCP Playground?

Azure Container Apps - AI & MCP Playground is a demo terminal application that showcases the Model Context Protocol (MCP) with OpenAI, Azure OpenAI, and GitHub Models. It provides an agent that uses tools from an MCP server to manage a TODO list, and is intended for developers exploring MCP-based agentic workflows.

How to use Azure Container Apps - AI & MCP Playground?

Clone the repository, install dependencies with npm install, configure your LLM provider (OpenAI API key, Azure OpenAI endpoint, or GitHub token) in a .env file, then start the MCP servers (HTTP and SSE) using docker compose up or npm start in separate terminals. Finally, run the MCP host with npm start --prefix mcp-host and interact with the agent through a terminal or VS Code’s built-in MCP support.

Key features of Azure Container Apps - AI & MCP Playground

Two MCP server implementations: HTTP Streaming and SSE (legacy)
Tools to add, list, complete, and delete TODO items
Persistent state via DocumentDB Local database
Support for OpenAI, Azure OpenAI, and GitHub Models as LLM providers
Debug mode via DEBUG=mcp:* environment variable
Runs locally with Docker or directly on Node.js 22+

Use cases of Azure Container Apps - AI & MCP Playground

Learn and prototype MCP-based agent architectures
Test tool calling across multiple LLM providers (OpenAI, Azure OpenAI, GitHub Models)
Demonstrate a TODO list agent with persistent state in a local database
Experiment with both SSE and HTTP streaming MCP transports

FAQ from Azure Container Apps - AI & MCP Playground

What LLM providers are supported?

OpenAI (Responses API), Azure OpenAI (Responses API), and GitHub Models (ChatCompletion API) are supported. Authentication uses API keys or, for Azure OpenAI, Managed Identity (not available in local Docker containers).

How do I set up the MCP servers?

Two servers are provided: mcp-server-http and mcp-server-sse. They can be started via Docker Compose (docker compose up) or manually with npm start --prefix mcp-server-http and npm start --prefix mcp-server-sse in separate terminals.

Where is data stored?

All agent state and tool data (e.g., TODO items) are persisted in a DocumentDB Local database running in Docker. You can explore it using the VS Code DocumentDB extension.

What transport protocols are supported?

The MCP server supports both SSE (Server-Sent Events) and HTTP Streaming transports. The MCP host scans for both servers and uses whichever is available.

Is authentication supported?

Token-based authentication is a work‑in‑progress (wip). When running locally in Docker, Azure OpenAI access via Managed Identity is not supported; you must use an API key instead.