Ir para o conteúdo

Lab 028: Implantar Servidor MCP no Azure Container Apps

Nível: L300 Trilha: MCP Tempo: ~60 min 💰 Custo: ~$0 (scale-to-zero) + Azure Container Registry ~$5/month

Assinatura Azure necessária

Este lab implanta Azure Container Apps. → Guia de pré-requisitos

O que Você Vai Aprender

  • Containerizar um servidor MCP (Python) com Docker
  • Enviar a imagem para o Azure Container Registry
  • Implantar no Azure Container Apps com transporte SSE
  • Atualizar o servidor MCP com deploys de atualização contínua sem tempo de inatividade
  • Conectar a partir do GitHub Copilot e qualquer cliente MCP

Pré-requisitos

Implantar Infraestrutura

Opção A — Implantar no Azure (um clique)

Deploy to Azure

Isto implanta: - Ambiente Azure Container Apps + Log Analytics - Um container app placeholder (você o atualizará com sua imagem no Passo 3)

Opção B — Azure CLI (Bicep)

git clone https://github.com/lcarli/AI-LearningHub.git && cd AI-LearningHub

az group create --name rg-ai-labs-mcp --location eastus

az deployment group create \
  --resource-group rg-ai-labs-mcp \
  --template-file infra/lab-028-mcp-container-apps/main.bicep

# Get the app URL
az deployment group show \
  --resource-group rg-ai-labs-mcp \
  --name main \
  --query properties.outputs.appUrl.value -o tsv

Exercício do Lab

Passo 1: Criar o servidor MCP (transporte SSE)

Para implantação remota, use transporte SSE (Server-Sent Events) em vez de stdio.

mcp-product-server/
├── server.py
├── Dockerfile
└── requirements.txt

server.py:

import os, csv, json
from mcp.server.fastmcp import FastMCP

mcp = FastMCP("OutdoorGear Product Server")

# Load product catalog from GitHub raw URL or local file
PRODUCTS_URL = "https://raw.githubusercontent.com/lcarli/AI-LearningHub/main/data/products.csv"

import urllib.request, io

def load_products() -> list[dict]:
    with urllib.request.urlopen(PRODUCTS_URL) as r:
        content = r.read().decode("utf-8")
    return list(csv.DictReader(io.StringIO(content)))

PRODUCTS = load_products()

@mcp.tool()
def search_products(query: str) -> str:
    """Search outdoor gear products by keyword."""
    q = query.lower()
    matches = [
        p for p in PRODUCTS
        if q in p["name"].lower() or q in p["category"].lower() or q in p["description"].lower()
    ]
    if not matches:
        return "No products found."
    return json.dumps(matches[:5], indent=2)

@mcp.tool()
def get_product(sku: str) -> str:
    """Get full details for a product by SKU."""
    for p in PRODUCTS:
        if p["sku"].lower() == sku.lower():
            return json.dumps(p, indent=2)
    return f"Product with SKU '{sku}' not found."

@mcp.tool()
def list_categories() -> str:
    """List all available product categories."""
    cats = sorted(set(f"{p['category']}/{p['subcategory']}" for p in PRODUCTS))
    return "\n".join(cats)

if __name__ == "__main__":
    # Use SSE transport for remote/cloud deployment
    mcp.run(transport="sse", host="0.0.0.0", port=int(os.environ.get("PORT", 8000)))

requirements.txt: 

mcp[cli]>=1.0.0

Dockerfile: 

FROM python:3.12-slim
WORKDIR /app
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt
COPY server.py .
EXPOSE 8000
CMD ["python", "server.py"]

Passo 2: Testar localmente

cd mcp-product-server
pip install mcp
python server.py
# Server running on http://0.0.0.0:8000

# Test in another terminal
npx @modelcontextprotocol/inspector http://localhost:8000/sse

Passo 3: Construir e enviar a imagem do container

# Variables — update these
RESOURCE_GROUP="rg-ai-labs-mcp"
ACR_NAME="mcpacr$(date +%s | tail -c 6)"   # unique name

# Create Azure Container Registry (Basic tier ~$5/month)
az acr create --resource-group $RESOURCE_GROUP --name $ACR_NAME --sku Basic

# Build and push
az acr build --registry $ACR_NAME --image mcp-product-server:latest .

# Get the image URL
IMAGE_URL="${ACR_NAME}.azurecr.io/mcp-product-server:latest"
echo "Image: $IMAGE_URL"

Passo 4: Implantar no Container Apps

# Get the app name from deployment outputs
APP_NAME=$(az deployment group show \
  --resource-group $RESOURCE_GROUP \
  --name main \
  --query properties.outputs.appName.value -o tsv)

ENV_NAME=$(az deployment group show \
  --resource-group $RESOURCE_GROUP \
  --name main \
  --query properties.outputs.environmentName.value -o tsv)

# Enable Container Apps to pull from ACR
az containerapp registry set \
  --name $APP_NAME \
  --resource-group $RESOURCE_GROUP \
  --server "${ACR_NAME}.azurecr.io" \
  --identity system

# Update the container app with your image
az containerapp update \
  --name $APP_NAME \
  --resource-group $RESOURCE_GROUP \
  --image $IMAGE_URL

# Get the public URL
az containerapp show \
  --name $APP_NAME \
  --resource-group $RESOURCE_GROUP \
  --query properties.configuration.ingress.fqdn -o tsv

Passo 5: Conectar a partir do GitHub Copilot

Add to .vscode/mcp.json:

{
  "servers": {
    "outdoorgear-cloud": {
      "type": "sse",
      "url": "https://YOUR-APP-NAME.YOUR-REGION.azurecontainerapps.io/sse"
    }
  }
}

Teste no Copilot Agent Mode: "Quais barracas de camping vocês têm em estoque?"

Passo 6: Atualizações sem tempo de inatividade

# Update to a new version
az acr build --registry $ACR_NAME --image mcp-product-server:v2 .

az containerapp update \
  --name $APP_NAME \
  --resource-group $RESOURCE_GROUP \
  --image "${ACR_NAME}.azurecr.io/mcp-product-server:v2"

# Container Apps performs a rolling update automatically

Detalhamento de Custos

Recurso SKU Custo Estimado
Container Apps Scale-to-zero (0 replicas when idle) ~$0 when idle
Container Apps (active) 0.5 vCPU / 1Gi ~$0.02/hour active
Azure Container Registry Basic ~$5/month
Log Analytics PerGB2018 ~$2/month (light usage)

Escalar até zero

O Container Apps escala para 0 réplicas quando não há requisições. Para um servidor MCP de dev/lab com uso ocasional, você pagará quase nada.

Limpeza

az group delete --name rg-ai-labs-mcp --yes --no-wait

Próximos Passos