eron1703 Claude Pod - Remote Claude Code Installation
Quick Access
One-Click Connection (Recommended):
- Mac: Double-click
connect-claude-pod.command - Windows: Double-click
connect-claude-pod.bat
Manual SSH Connection:
ssh dev-01-root "kubectl exec -it -n claude-pod deployment/claude-pod -- bash"
Setup (One-Time)
Place SSH private key at:
- Mac/Linux:
~/.ssh/id_ed25519_demo - Windows:
C:\Users\YourUsername\.ssh\id_ed25519_demo
Get key from:
glab variable get SSH_PRIVATE_KEY --group flow-master
What's Inside
- Claude Code 2.1.37 - Full CLI with all features
- test-rig 1.0.0 - Testing orchestration framework
- 40 Skills - Auto-loaded from claude-config-loader
- LiteLLM Proxy - AI model routing for cost savings
Model Routing (Cost Optimization)
claude-3-5-sonnet → Grok-3 ($0.70/1M tokens) # 77% cheaper
claude-3-opus → Gemini 2.5 Flash ($3.50/1M) # 12% cheaper
claude-haiku → Gemini 2.5 Flash ($3.50/1M) # Actually more expensive!
Cost Tracking:
# Inside pod
curl http://litellm-proxy:4000/spend/tags
Deployment Info
- Server: dev-01 (65.21.153.235)
- Namespace:
claude-pod - Platform: K3S (Kubernetes)
- Replicas: 1-8 (auto-scales with HPA)
- Resources: 500m-2000m CPU, 1Gi-4Gi memory per pod
- Deployment: Auto-deploy on push to main (GitLab CI/CD)
Health Check
# From dev-01
kubectl get pods -n claude-pod
kubectl logs -f -n claude-pod deployment/claude-pod
# From pod
claude --version # Should show 2.1.37
test-rig --version # Should show 1.0.0
ls ~/.claude/skills/ # Should show 40 skills
Troubleshooting
Pod not responding:
kubectl rollout restart deployment/claude-pod -n claude-pod
Skills not loaded:
kubectl delete pod -n claude-pod -l app.kubernetes.io/name=claude-pod
# Wait for pod to restart (init container will reload skills)
LiteLLM not routing:
# Inside pod
curl http://litellm-proxy:4000/health
# Check settings
cat ~/.claude/settings.json | grep baseUrl
Repository
- GitLab: https://gitlab.com/flow-master/claude-pod
- CI/CD: Auto-deploy on main branch
- Monitoring: Grafana dashboard (planned)