eks-sage-enterprise 2.0.1

Enterprise-grade MCP server for Amazon EKS — 62 intelligent tools across 12 categories

EKS Sage Enterprise

62 MCP tools that turn 3-hour EKS incidents into 10-second diagnoses.

---

How It Works

eks-sage runs locally on your machine and connects to your own AWS account using your existing AWS credentials. You keep full control — no data leaves your machine except your own AWS API calls.

Your Machine
├── ~/.aws/credentials     ← your AWS keys (already configured)
├── ~/.kube/config         ← your cluster config (updated by eks-sage)
└── eks-sage-enterprise    ← reads credentials, talks to YOUR AWS
        ↓
YOUR AWS Account
        ↓
YOUR EKS Cluster

---

Prerequisites

Before installing, make sure you have these on your machine:

1 — AWS CLI + profiles configured

# Install AWS CLI
brew install awscli

# Check your existing profiles
cat ~/.aws/config

# You should have profiles like:
# [profile dev]
# [profile nonprod]
# [profile prod]

# Verify a profile works
aws sts get-caller-identity --profile dev
aws sts get-caller-identity --profile nonprod
aws sts get-caller-identity --profile prod

Your ~/.aws/config typically looks like:

[profile dev]
region = us-east-1
role_arn = arn:aws:iam::111111111111:role/DevRole
source_profile = default

[profile nonprod]
region = us-east-1
role_arn = arn:aws:iam::222222222222:role/NonProdRole
source_profile = default

[profile prod]
region = us-east-1
role_arn = arn:aws:iam::333333333333:role/ProdRole
source_profile = default

Then tell eks-sage which profile to use:

# For dev cluster
AWS_PROFILE=dev claude mcp add eks-sage -s user -- python -m eks_sage_enterprise.server

# Or set it in your environment
export AWS_PROFILE=nonprod

Or just ask Claude:

select profile nonprod
connect to cluster my-nonprod-cluster in us-east-1

2 — kubectl

brew install kubectl

3 — Python 3.11+

python3 --version
# Should show 3.11 or higher

4 — A Claude MCP client

Works with: Claude Desktop · Claude Code · Cursor · Windsurf · Cline · Zed

---

Install

pip install eks-sage-enterprise
claude mcp add eks-sage -s user -- python -m eks_sage_enterprise.server

Verify it connected:

claude mcp list
# Should show: eks-sage ✅ Connected

---

Quick Start

# Connect to your cluster first
connect to cluster my-cluster in us-east-1

# Then ask anything
give me a full cluster health check
investigate NLB service my-api user_ip 203.0.113.45
investigate IRSA for service account my-sa in namespace production
check compliance profile cis

---

Your Data stays on Your Machine

eks-sage ONLY makes calls to:
  → AWS APIs (*.amazonaws.com) using YOUR credentials
  → Your kubectl (~/.kube/config)
  → CloudWatch logs in YOUR account

eks-sage NEVER:
  → Sends data to any third party
  → Stores your credentials
  → Shares anything outside your machine

62 Tools · 12 Categories

Category	Tools	Description
Guardrails	5	Safety mode, confirmation gate, audit log
Cluster	5	List, describe, connect, addons, upgrade insights
Nodes	5	Nodegroups, nodes, usage, events, cordon
Workloads	8	Pods, logs, deployments, daemonsets, statefulsets, jobs, cronjobs
Security	7	IRSA, RBAC audit, pod security, secrets, IAM mapping, service accounts
Networking	7	Services, ingresses, namespaces, DNS check, NLB investigation
Troubleshooting	8	Investigate pod/daemonset/statefulset/cronjob, health check, incident summary
Storage	3	PVs, PVCs, storage investigation
Scaling	4	HPA, quotas, PDBs, cost by namespace
Observability	3	CloudWatch, Container Insights, log aggregation
Multi-Cluster	3	Fleet view, compare, switch context
Compliance	4	CIS benchmark, drift, deprecations, audit trail

---

What to Ask — All 12 Categories

🛡️ Guardrails

set safety mode to standard
get safety status
confirm A1B2C3D4
show me the audit log

🏗️ Cluster Management

list my EKS clusters in us-east-1
connect to cluster my-cluster in us-east-1
what addons are installed on my cluster
is my cluster ready to upgrade
describe cluster my-cluster

🖥️ Node Management

show all nodes and their status
show CPU and memory usage per node
list my node groups
get events for node i-0abc123

📦 Workload Operations

show all pods across namespaces
get logs from pod my-app-xyz in namespace production
list all deployments
show all daemonsets
list statefulsets in namespace data
show CPU usage per pod
show jobs and cronjobs

🔐 Security & RBAC

investigate IRSA for service account my-sa in namespace production
audit RBAC permissions
check pod security across all namespaces
scan for secrets exposed as env vars
who has access to my cluster
show IAM to Kubernetes permission mapping

🌐 Networking & NLB

show all services and load balancers
investigate NLB service my-api namespace production user_ip 1.2.3.4
investigate NLB service my-api user_domain app.example.com
list all ingresses
check DNS resolution for my-service in namespace production
show network policies
list configmaps and secrets in namespace production

🔍 Troubleshooting & Incidents ← Most Used

give me a full cluster health check
something is broken — give me a full incident report
investigate pod my-app-xyz in namespace production
check for node pressure and OOM evictions
show all warning events in the cluster
investigate daemonset aws-node in kube-system
investigate statefulset my-db in namespace data
investigate cronjob my-backup in namespace ops

💾 Storage

show all PVs and PVCs
investigate stuck PVC my-data-pvc in namespace production
list storage classes

📈 Scaling & Cost

show horizontal pod autoscalers
show resource quotas per namespace
show pod disruption budgets
get cost breakdown by namespace

👁️ Observability

get CloudWatch metrics for my cluster
search Container Insights for ERROR logs last 2 hours
get logs from all pods with label app=my-api filter ERROR
show cost breakdown by namespace

🌍 Multi-Cluster

list all my EKS clusters across us-east-1 and eu-west-1
compare cluster staging to cluster production
switch to production cluster in us-west-2

✅ Compliance & Drift

check compliance profile cis
detect drift in my cluster last 24 hours
check for deprecated APIs before upgrading
show all cluster changes in last 24 hours

---

5-Layer Guardrail System

Layer 1 — Operation Classification  (READ/CONFIG/WRITE/DESTRUCTIVE)
Layer 2 — Safety Mode               (read_only / standard / unrestricted)
Layer 3 — Denylist                  (74 permanently blocked operations)
Layer 4 — Confirmation Gate         (writes require explicit approval)
Layer 5 — Audit Log                 (/tmp/eks-sage-audit.log)

License

MIT — free for everyone. Commercial use welcome.