Cloud Deployment Guide

This article provides a comprehensive step-by-step guide for deploying the entire AI agent system to Google Cloud Platform.

Prerequisites

Google Cloud Platform account
gcloud CLI installed and configured
Docker installed locally
Firebase project created
Service account with appropriate permissions

Step 1: Set Up Firebase

1.1 Create Firebase Project

# Create Firebase project
firebase projects:create <PROJECT_ID>

# Set as default project
firebase use <PROJECT_ID>

1.2 Enable Firebase Authentication

Go to Firebase Console → Authentication
Enable Email/Password authentication
Configure authorized domains

1.3 Create Firestore Database

# Initialize Firestore
firebase init firestore

# Create database in production mode
# Select region (e.g., australia-southeast1)

1.4 Create Service Account

Go to Firebase Console → Project Settings → Service Accounts
Click “Generate New Private Key”
Save the JSON file securely
Upload to Secret Manager (see Step 2)

Step 2: Set Up Secret Manager

2.1 Create Secret

# Create secret for Firebase service account
gcloud secrets create firebase-service-account \
  --data-file=./firebase-service-account.json \
  --project=<PROJECT_ID>

2.2 Grant Access

# Grant Cloud Run access to secret
gcloud secrets add-iam-policy-binding firebase-service-account \
  --member="serviceAccount:PROJECT_NUMBER-compute@developer.gserviceaccount.com" \
  --role="roles/secretmanager.secretAccessor" \
  --project=<PROJECT_ID>

Step 3: Build and Push Docker Images

3.1 Configure Docker for GCP

# Configure Docker to use gcloud as credential helper
gcloud auth configure-docker

3.2 Build Images

# Build Product Catalogue MCP
cd mcp/mcp-product-catalogue
docker build -t gcr.io/<PROJECT_ID>/mcp-product-catalogue:latest .

# Build CRM MCP
cd ../mcp-crm
docker build -t gcr.io/<PROJECT_ID>/mcp-crm:latest .

# Build Support Agent
cd ../../agents/support-agent-adk
docker build -t gcr.io/<PROJECT_ID>/support-agent-adk:latest .

3.3 Push Images

# Push all images
docker push gcr.io/<PROJECT_ID>/mcp-product-catalogue:latest
docker push gcr.io/<PROJECT_ID>/mcp-crm:latest
docker push gcr.io/<PROJECT_ID>/support-agent-adk:latest

Step 4: Deploy MCP Servers

4.1 Deploy Product Catalogue MCP

gcloud run deploy mcp-product-catalogue \
  --image gcr.io/<PROJECT_ID>/mcp-product-catalogue:latest \
  --platform managed \
  --region australia-southeast1 \
  --allow-unauthenticated \
  --set-env-vars FIREBASE_PROJECT_ID=<PROJECT_ID> \
  --set-env-vars GOOGLE_CLOUD_PROJECT=<PROJECT_ID> \
  --set-secrets FIREBASE_SERVICE_ACCOUNT=firebase-service-account:latest \
  --memory 512Mi \
  --cpu 1 \
  --min-instances 0 \
  --max-instances 10

Note the service URL from the output (e.g., https://mcp-product-catalogue-xxx.run.app)

4.2 Deploy CRM MCP

gcloud run deploy mcp-crm \
  --image gcr.io/<PROJECT_ID>/mcp-crm:latest \
  --platform managed \
  --region australia-southeast1 \
  --allow-unauthenticated \
  --set-env-vars FIREBASE_PROJECT_ID=<PROJECT_ID> \
  --set-env-vars GOOGLE_CLOUD_PROJECT=<PROJECT_ID> \
  --set-secrets FIREBASE_SERVICE_ACCOUNT=firebase-service-account:latest \
  --memory 512Mi \
  --cpu 1 \
  --min-instances 0 \
  --max-instances 10

Note the service URL from the output (e.g., https://mcp-crm-xxx.run.app)

Step 5: Deploy Support Agent

gcloud run deploy agent-order-receiver \
  --image gcr.io/<PROJECT_ID>/support-agent-adk:latest \
  --platform managed \
  --region australia-southeast1 \
  --allow-unauthenticated \
  --set-env-vars FIREBASE_PROJECT_ID=<PROJECT_ID> \
  --set-env-vars CATALOGUE_MCP_URL=https://mcp-product-catalogue-xxx.run.app \
  --set-env-vars CRM_MCP_URL=https://mcp-crm-xxx.run.app \
  --set-env-vars CORS_ORIGINS=https://<PROJECT_ID>.web.app,https://<PROJECT_ID>.firebaseapp.com \
  --set-secrets FIREBASE_SERVICE_ACCOUNT=firebase-service-account:latest \
  --memory 1Gi \
  --cpu 2 \
  --min-instances 0 \
  --max-instances 10

Note the service URL from the output (e.g., https://agent-order-receiver-xxx.run.app)

Step 6: Configure CORS

CORS is automatically configured via environment variables. Verify:

# Check Support Agent CORS
curl -H "Origin: https://<PROJECT_ID>.web.app" \
  -H "Access-Control-Request-Method: POST" \
  -H "Access-Control-Request-Headers: authorization" \
  -X OPTIONS \
  https://agent-order-receiver-xxx.run.app/process

Step 7: Set Up Cloud Storage (for LanceDB)

7.1 Create Bucket

# Create bucket for LanceDB data
gsutil mb -p <PROJECT_ID> -c STANDARD -l australia-southeast1 gs://<PROJECT_ID>-lancedb

7.2 Upload Data

# Upload LanceDB data
gsutil -m cp -r ./data/lancedb/* gs://<PROJECT_ID>-lancedb/

Step 8: Deploy Frontend (Optional)

8.1 Build Next.js App

cd frontend
npm run build

8.2 Deploy to Firebase Hosting

firebase deploy --only hosting

Step 9: Test Deployment

9.1 Health Checks

# Test Support Agent
curl https://agent-order-receiver-xxx.run.app/health

# Test Product Catalogue MCP
curl https://mcp-product-catalogue-xxx.run.app/health

# Test CRM MCP
curl https://mcp-crm-xxx.run.app/health

9.2 Test Authentication Flow

# Get Firebase token (from frontend)
FIREBASE_TOKEN="your-firebase-token"

# Test message processing
curl -X POST https://agent-order-receiver-xxx.run.app/process \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $FIREBASE_TOKEN" \
  -d '{
    "message": "Do you sell organic products?",
    "conversation_history": []
  }'

9.3 Test MCP Calls

# Get service account token (from Support Agent logs)
SERVICE_TOKEN="service-account-token"

# Test Product Catalogue MCP
curl -X POST https://mcp-product-catalogue-xxx.run.app/mcp \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer $SERVICE_TOKEN" \
  -d '{
    "jsonrpc": "2.0",
    "method": "tools/call",
    "params": {
      "name": "search_products",
      "arguments": {
        "query": "organic products",
        "limit": 10
      }
    },
    "id": 1
  }'

Step 10: Set Up Monitoring

10.1 Enable Cloud Monitoring

# Enable Cloud Monitoring API
gcloud services enable monitoring.googleapis.com

10.2 Create Alert Policies

# Create alert for high error rate
gcloud alpha monitoring policies create \
  --notification-channels=CHANNEL_ID \
  --display-name="High Error Rate" \
  --condition-threshold-value=0.05 \
  --condition-threshold-duration=300s

10.3 Set Up Logging

# View logs
gcloud logging read "resource.type=cloud_run_revision" --limit 50

# Export logs
gcloud logging export logs.json \
  --log-filter='resource.type=cloud_run_revision'

Troubleshooting

Common Issues

Service Not Starting
- Check Cloud Run logs: gcloud run services logs read SERVICE_NAME
- Verify environment variables
- Check secret access permissions
Authentication Failures
- Verify Firebase project ID matches
- Check service account key is valid
- Verify token expiration
CORS Errors
- Check CORS_ORIGINS environment variable
- Verify frontend URL is included
- Check Cloud Run CORS configuration
MCP Call Failures
- Verify MCP URLs are correct
- Check service account token is valid
- Verify network connectivity

Cost Optimization

Cloud Run: Use min instances = 0 for cost savings
Firestore: Use on-demand pricing for low traffic
Cloud Storage: Use standard storage class
Monitoring: Set up budget alerts

Security Checklist

Secrets stored in Secret Manager
Service accounts have minimal permissions
CORS configured correctly
HTTPS enforced on all endpoints
Token validation implemented
Error messages don’t leak sensitive info

Next Steps

In the final article, we’ll cover Monitoring, Security & Maintenance, including observability, troubleshooting, and best practices for maintaining production AI agents.