DeepGanguly presents a step-by-step guide to hosting a Model Context Protocol (MCP) server with HTTP transport on Azure Container Apps, enabling integration with external APIs and use in tools like VS Code Chat.

Hosting Remote MCP Server on Azure Container Apps Using HTTP Transport

About

Continuing from a previous exploration of using Server-Sent Events (SSE) transports in Azure Container Apps (ACA), this article focuses on hosting Remote Model Context Protocol (MCP) Servers in ACA implementing the HTTP transport mechanism.

Overview

The Model Context Protocol (MCP) changes how AI assistants interface with external data and tools. While local stdio deployments are common, this guide shows how to create and deploy a production-grade MCP server utilizing HTTP transport in Azure Container Apps. As an example, a live forex currency converter server is built that retrieves real-time rates from external APIs, demonstrating how AI assistants can benefit from up-to-date dynamic data via MCP integration.

What is MCP HTTP Transport?

MCP supports several transport mechanisms:

Stdio Transport: For direct process communication
HTTP Transport: For RESTful API communication, ideal for cloud scenarios

HTTP transport allows efficient cloud deployment, cross-platform compatibility, handling multiple clients, load balancing, third-party API integrations, and real-time external data fetching.

Building and Testing the MCP Server

Clone the Repository

git clone https://github.com/deepganguly/azure-container-apps-mcp-sample.git
cd azure-container-apps-mcp-sample
npm install

Run the MCP Server Locally
```
npm start
```

Test the HTTP Endpoint

curl -X POST http://localhost:3001/mcp \
 -H "Content-Type: application/json" \
 -d '{"jsonrpc":"2.0","id":1,"method":"tools/list","params":{}}'

Server Features

Exchange Rate Management: Caches rates from exchangerate-api.com for 10 minutes, updates as needed, and falls back to hardcoded values on API failures.
HTTP Endpoints: Listens for JSON-RPC 2.0 POST requests at /mcp.
- tools/list lists available tools like convert_currency.
- tools/call handles currency conversion requests.
Currency Conversion Logic: Uses USD as a base, supports cross-currency calculation via USD, and manages errors gracefully.
Responses: Returns formatted, informative responses including the result, live rate, and error handling.

Deploying to Azure Container Apps

Use the included Bicep template for one-step deployment or deploy manually with CLI commands:

# Login to Azure

az login

# Create a resource group

az group create --name mcp-live-rates-rg --location eastus

# Create a Container App environment

az containerapp env create --name mcp-forex-env --resource-group mcp-live-rates-rg --location eastus

# Deploy the container app

az containerapp up --name mcp-live-forex-server --resource-group mcp-live-rates-rg --environment mcp-forex-env --source . --target-port 3001 --ingress external

Connecting MCP Server to VS Code Chat

Get the MCP Server URL:
- Example: https://mcp-live-forex-server.<id>.eastus.azurecontainerapps.io/mcp

Configure VS Code:

Edit or create .vscode/mcp.json:

{
  "servers": {
    "my-mcp-server": {
      "url": "https://mcp-live-forex-server.<id>.eastus.azurecontainerapps.io/mcp",
      "type": "http"
    }
  },
  "inputs": [
    {
      "id": "convert_currency",
      "type": "promptString",
      "description": "Convert {amount} {from} to {to} using live exchange rates"
    }
  ]
}

Add your server, reload VS Code, and interact with the MCP server via chat.

Conclusion

By building and deploying a MCP HTTP server on Azure Container Apps, developers gain a robust, scalable, and flexible environment for cloud-based integrations with AI assistants. ACA offers operational benefits like easy monitoring, scaling, and updating through Azure tools and CLI.

Repository: azure-container-apps-mcp-sample

For further details and source code, visit the linked repository.

This post appeared first on “Microsoft Tech Community”. Read the entire article here