Deploy to GCP Cloud Run

Deploy Clinical Terminology MCP servers to Google Cloud Run.

Deploy MCP servers to Google Cloud Run for fully managed serverless containers.

Prerequisites

Google Cloud CLI (gcloud) installed and configured
Docker installed locally
GCP project with billing enabled

Set Up Project

# Set project
export PROJECT_ID=your-project-id
gcloud config set project $PROJECT_ID

# Enable required APIs
gcloud services enable \
  run.googleapis.com \
  artifactregistry.googleapis.com \
  secretmanager.googleapis.com

Create Artifact Registry Repository

gcloud artifacts repositories create clinical-terminology-mcp \
  --repository-format=docker \
  --location=us-central1 \
  --description="Clinical Terminology MCP server images"

# Configure Docker authentication
gcloud auth configure-docker us-central1-docker.pkg.dev

Build and Push Image

# Build image
docker build -t snomed-mcp -f snomed-mcp/Dockerfile .

# Tag for Artifact Registry
docker tag snomed-mcp \
  us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/snomed-mcp:latest

# Push
docker push \
  us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/snomed-mcp:latest

Alternatively, use Cloud Build:

gcloud builds submit \
  --tag us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/snomed-mcp:latest \
  --file snomed-mcp/Dockerfile \
  .

Store Secrets

For servers requiring credentials:

# ICD-11 credentials
echo -n "your-client-id" | gcloud secrets create icd11-client-id --data-file=-
echo -n "your-client-secret" | gcloud secrets create icd11-client-secret --data-file=-

# UMLS API key
echo -n "your-api-key" | gcloud secrets create umls-api-key --data-file=-

# LOINC credentials
echo -n "your-username" | gcloud secrets create loinc-username --data-file=-
echo -n "your-password" | gcloud secrets create loinc-password --data-file=-

Create Service Account

# Create service account
gcloud iam service-accounts create mcp-server \
  --display-name="MCP Server Service Account"

# Grant Secret Manager access
gcloud projects add-iam-policy-binding $PROJECT_ID \
  --member="serviceAccount:mcp-server@$PROJECT_ID.iam.gserviceaccount.com" \
  --role="roles/secretmanager.secretAccessor"

Deploy to Cloud Run

Basic Server (no secrets)

gcloud run deploy snomed-mcp \
  --image us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/snomed-mcp:latest \
  --platform managed \
  --region us-central1 \
  --port 8080 \
  --cpu 1 \
  --memory 512Mi \
  --min-instances 0 \
  --max-instances 10 \
  --set-env-vars "MCP_TRANSPORT=http,MCP_HTTP_ADDR=:8080,MCP_LOG_FORMAT=json,MCP_METRICS_ENABLED=true" \
  --allow-unauthenticated

Server with Secrets (e.g., ICD-11)

gcloud run deploy icd11-mcp \
  --image us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/icd11-mcp:latest \
  --platform managed \
  --region us-central1 \
  --port 8080 \
  --cpu 1 \
  --memory 512Mi \
  --min-instances 0 \
  --max-instances 10 \
  --service-account mcp-server@$PROJECT_ID.iam.gserviceaccount.com \
  --set-env-vars "MCP_TRANSPORT=http,MCP_HTTP_ADDR=:8080,MCP_LOG_FORMAT=json" \
  --set-secrets "ICD11_CLIENT_ID=icd11-client-id:latest,ICD11_CLIENT_SECRET=icd11-client-secret:latest" \
  --allow-unauthenticated

Configure Health Checks

Cloud Run automatically uses the container’s health check. The MCP servers expose:

/health/live - Liveness probe
/health/ready - Readiness probe

Cloud Run uses startup and liveness probes automatically.

Custom Domain (Optional)

# Map custom domain
gcloud run domain-mappings create \
  --service snomed-mcp \
  --domain mcp.yourdomain.com \
  --region us-central1

# View DNS records to configure
gcloud run domain-mappings describe \
  --domain mcp.yourdomain.com \
  --region us-central1

Verify Deployment

# Get service URL
SERVICE_URL=$(gcloud run services describe snomed-mcp \
  --region us-central1 \
  --format 'value(status.url)')

# Test health
curl $SERVICE_URL/health

# Test MCP endpoint
curl -X POST $SERVICE_URL/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","method":"initialize","params":{},"id":1}'

Deploy All Servers

Script to deploy all servers:

#!/bin/bash
SERVERS="snomed-mcp rxnorm-mcp icd10-mcp ucum-mcp"
SERVERS_WITH_SECRETS="icd11-mcp loinc-mcp umls-mcp"

for server in $SERVERS; do
  gcloud run deploy $server \
    --image us-central1-docker.pkg.dev/$PROJECT_ID/clinical-terminology-mcp/$server:latest \
    --platform managed \
    --region us-central1 \
    --port 8080 \
    --set-env-vars "MCP_TRANSPORT=http,MCP_HTTP_ADDR=:8080,MCP_LOG_FORMAT=json,MCP_METRICS_ENABLED=true" \
    --allow-unauthenticated
done

Monitoring

Cloud Run integrates with Cloud Monitoring. View metrics in the GCP Console:

Request count and latency
Container instance count
CPU and memory utilization

For Prometheus metrics, configure a Cloud Monitoring scrape job or use the managed Prometheus service.

Cost Optimization

Set --min-instances 0 to scale to zero when idle
Use --cpu-throttling to reduce costs during idle periods
Consider regional deployments to reduce latency and costs