Deployment Options

Deploy MCP Server with LangGraph to production

Deployment Options

Choose the deployment method that best fits your needs:

LangGraph Platform

One-command serverless deployment

• Best for: Quick production deployments
• Complexity: ★☆☆☆☆ Minimal
• Time: ~2 minutes*
• Cost: Pay-per-use

Google Cloud Run

GCP serverless container platform

• Best for: GCP-native applications
• Complexity: ★★☆☆☆ Low
• Time: ~10 minutes*
• Cost: Pay-per-use

Kubernetes

Production-grade K8s deployment

• Best for: Enterprise, self-hosted
• Complexity: ★★★★☆ High
• Time: ~1-2 hours*
• Cost: Fixed + usage

Docker

Quick Docker Compose setup

• Best for: Development, testing
• Complexity: ★☆☆☆☆ Minimal
• Time: ~15 minutes*
• Cost: Hosting only

Helm

Flexible Helm chart deployment

• Best for: Customizable K8s
• Complexity: ★★★☆☆ Medium
• Time: ~30 minutes*
• Cost: Fixed + usage

Kustomize

Environment-specific overlays

• Best for: Multi-environment K8s
• Complexity: ★★★☆☆ Medium
• Time: ~45 minutes*
• Cost: Fixed + usage

*Time estimates assume prerequisites are configured (accounts, credentials, CLI tools installed). First-time setup may take longer. See individual deployment guides for detailed requirements.

Platform Comparison

Feature	LangGraph Platform	Cloud Run	Kubernetes	Docker
Setup Time	~2 min*	~10 min*	~1-2 hrs*	~15 min*
Infrastructure	✅ None	⚠️ Minimal	❌ Complex	⚠️ Basic
Scaling	✅ Auto	✅ Auto	⚠️ Manual config	❌ Manual
LangSmith Integration	✅ Built-in	⚠️ Manual	⚠️ Manual	⚠️ Manual
Versioning	✅ Built-in	⚠️ Manual	⚠️ Manual	❌ None
Cost	Pay-per-use	Pay-per-use	Fixed + usage	Hosting only
Best For	Quick production	GCP apps	Enterprise	Development

*Time estimates with prerequisites configured. First-time setup takes longer.

Recommendation: Start with LangGraph Platform for fastest time-to-production, or Cloud Run if you’re already on GCP. Use Kubernetes for enterprise self-hosted deployments.

Architecture

Supported Platforms

Google Cloud

Google Kubernetes Engine (GKE)

Fully supported and tested

Features:

• ✅ Autopilot and Standard clusters
• ✅ Workload Identity for secrets
• ✅ Cloud Armor for DDoS protection
• ✅ Cloud Load Balancing
• ✅ Managed Prometheus

Deploy to GKE →

Pre-Deployment Checklist

1. Review Security Audit

   Run the security audit checklist:

   python scripts/validate_production.py --strict
   

   Ensure all checks pass before deploying.

2. Configure Secrets

   Set up secrets management:

   • Configure Infisical project
   • Generate JWT secret: openssl rand -base64 32
   • Get LLM API keys
   • Setup OpenFGA store and model IDs

   See Secrets Management

3. Build Container Image

   docker build -t your-registry/mcp-server-langgraph:1.0.0 .
   docker push your-registry/mcp-server-langgraph:1.0.0
   

   Or use GitHub Actions for automated builds.

4. Prepare Infrastructure

   • Provision Kubernetes cluster
   • Setup DNS records
   • Configure TLS certificates
   • Deploy monitoring stack
   • Setup OpenFGA with PostgreSQL backend
5. Review Configuration

   Update configuration for production:

   • Set ENVIRONMENT=production
   • Configure resource limits
   • Enable autoscaling
   • Setup network policies
   • Configure ingress and TLS

Quick Deploy (Docker)

For a quick test deployment:

## 1. Configure environment
cp .env.example .env.production
## Edit .env.production with your values

## 2. Start services
docker compose -f docker-compose.yml up -d

## 3. Verify
curl http://localhost:8000/health

Production Deploy (Kubernetes)

For production deployment:

## 1. Create namespace
kubectl create namespace mcp-server-langgraph

## 2. Create secrets
kubectl create secret generic mcp-server-langgraph-secrets \
  --from-literal=jwt-secret-key="$(openssl rand -base64 32)" \
  --from-literal=google-api-key="YOUR_KEY" \
  -n mcp-server-langgraph

## 3. Deploy with Helm
helm upgrade --install mcp-server-langgraph ./helm/mcp-server-langgraph \
  --namespace mcp-server-langgraph \
  --values values-production.yaml \
  --wait

## 4. Verify deployment
kubectl rollout status deployment/mcp-server-langgraph -n mcp-server-langgraph

Environment Configuration

Development

environment: development
replicaCount: 1
autoscaling:
  enabled: false
resources:
  limits:
    cpu: 500m
    memory: 512Mi

Monitoring & Observability

Deploy the monitoring stack:

## Prometheus + Grafana
helm install prometheus prometheus-community/kube-prometheus-stack \
  --namespace monitoring --create-namespace

## Jaeger
helm install jaeger jaegertracing/jaeger \
  --namespace monitoring

## OpenTelemetry Collector
kubectl apply -f kubernetes/otel-collector/

Access dashboards:

• Grafana: http://grafana.yourdomain.com (admin/admin)
• Jaeger: http://jaeger.yourdomain.com
• Prometheus: http://prometheus.yourdomain.com

Scaling

Horizontal Pod Autoscaling

apiVersion: autoscaling/v2
kind: HorizontalPodAutoscaler
metadata:
  name: mcp-server-langgraph
spec:
  scaleTargetRef:
    apiVersion: apps/v1
    kind: Deployment
    name: mcp-server-langgraph
  minReplicas: 3
  maxReplicas: 10
  metrics:
  - type: Resource
    resource:
      name: cpu
      target:
        type: Utilization
        averageUtilization: 70
  - type: Resource
    resource:
      name: memory
      target:
        type: Utilization
        averageUtilization: 80

Vertical Pod Autoscaling

kubectl apply -f kubernetes/vpa/

High Availability

Ensure high availability with:

1. Multiple Replicas: Run at least 3 replicas
2. Pod Disruption Budget: Maintain minimum availability
3. Multi-Zone Deployment: Spread across availability zones
4. Health Checks: Liveness and readiness probes
5. Graceful Shutdown: Handle SIGTERM properly
6. Circuit Breakers: Fail fast with timeouts

Disaster Recovery

1. Backup Strategy

   • OpenFGA PostgreSQL database
   • Configuration and secrets
   • Persistent volumes
2. Recovery Procedure

   • Restore database from backup
   • Redeploy application
   • Verify functionality
3. Testing

   • Test disaster recovery quarterly
   • Document RTO/RPO
   • Update runbooks

Security Hardening

Production security checklist:

• TLS enabled for all endpoints
• Network policies applied
• Pod security policies enforced
• RBAC configured with least privilege
• Secrets encrypted at rest
• Audit logging enabled
• Regular security scans
• Rate limiting configured

See Security Best Practices

Cost Optimization

GKE

• Use Autopilot for hands-off management
• Enable cluster autoscaling
• Use preemptible nodes for dev/staging
• Configure resource requests accurately
• Use committed use discounts

Next Steps

Docker Deployment

Quick deployment guide

Kubernetes Setup

Production Kubernetes deployment

Monitoring Setup

Configure observability

Security Guide

Production security hardening