Kubernetes MCP Server - Manage K8s/OpenShift via MCP¶

Repository: Manusa/kubernetes-mcp-server Stars: 1,814 ★ Category: DevOps / Infrastructure Language: Go License: Apache 2.0 Last Updated: Active (July 2026)

What It Does¶

A versatile Model Context Protocol server for Kubernetes and OpenShift. Distributed as a native binary, npm package, Python package, or Docker image. AI agents can interact with Kubernetes clusters - list pods, create deployments, check service status, view logs - all through MCP tools without kubectl CLI access.

Key Capabilities¶

Capability	Description
Cluster Management	List nodes, check cluster health, view resource utilization
Pod Operations	List pods, get pod details, view logs, exec into containers
Deployment Management	Create, update, scale, and rollback deployments
Service & Ingress	List services, check endpoints, view ingress rules
Config & Secrets	Read ConfigMaps and Secrets (read-only recommended)
Namespace Operations	List namespaces, create/delete (with guardrails)
OpenShift Support	Full OpenShift compatibility including Routes, BuildConfigs, DeploymentConfigs

Why Business Operators Care¶

Infrastructure visibility: Non-technical operators can query cluster state via natural language - "how many pods are running in production?"
Deployment auditing: Track what's deployed, when, and by whom without kubectl access
Incident response: AI agent can surface pod failures, crash loops, and resource exhaustion before humans notice
Multi-cluster: Single MCP server can connect to multiple clusters via kubeconfig contexts

Setup for Hermes Agent¶

Prerequisites¶

Access to a Kubernetes cluster (kubeconfig file)
kubectl or oc (for OpenShift) CLI authenticated

Option 1: Native Binary (Recommended)¶

# Download the latest binary for your architecture
curl -L "https://github.com/Manusa/kubernetes-mcp-server/releases/latest/download/kubernetes-mcp-server-linux-arm64" -o /usr/local/bin/kubernetes-mcp-server
chmod +x /usr/local/bin/kubernetes-mcp-server

# Verify
kubernetes-mcp-server --version

Option 2: Docker¶

docker run -d \
  -v ~/.kube/config:/root/.kube/config:ro \
  -p 8080:8080 \
  ghcr.io/manusa/kubernetes-mcp-server:latest

MCP Configuration¶

Add to ~/.hermes/config.yaml:

mcp_servers:
  kubernetes:
    type: stdio
    command: kubernetes-mcp-server
    args:
      - --kubeconfig
      - ~/.kube/config
      - --read-only  # Recommended for AI agent access
    env:
      KUBECONFIG: ~/.kube/config

Security Recommendations¶

Use read-only mode for AI agent access - prevents accidental deletion
Dedicated service account with minimal RBAC permissions
Namespace isolation - scope the kubeconfig context to non-production namespaces first
Audit logging - all MCP actions are logged; forward to SIEM if in production

Comparison: Kubernetes MCP vs CLI vs Dashboard¶

Feature	K8s MCP Server	kubectl CLI	K8s Dashboard
Access method	MCP tools (AI-native)	Terminal commands	Web UI
Natural language	Yes - AI translates	No	No
Automation	AI agent driven	Script-driven	Manual
Multi-cluster	Single config	Switch contexts	Per-dashboard
Audit trail	MCP logs	Shell history	Dashboard logs
Best for	AI-assisted operations, incident response	DevOps engineers	Visual monitoring

Common Queries via Hermes¶

Once configured, ask Hermes:

"Show me all failing pods in the production namespace"
"What deployments were updated in the last 24 hours?"
"Scale the api-gateway deployment to 5 replicas" (requires write mode)
"Get logs from the payment-service pod for the last hour"
"Are there any nodes running low on memory?"

Pitfalls¶

RBAC scope: AI agents inherit the kubeconfig permissions. A cluster-admin kubeconfig gives the AI full cluster control - use dedicated service accounts.
Rate limiting: Large clusters with thousands of pods can produce verbose output. Use namespace and label filters.
OpenShift vs vanilla K8s: Some tools (Routes, BuildConfigs) only work on OpenShift. Check API compatibility.

Discovered July 16, 2026 during mcpservers.org sitemap scan. 1,814 stars at time of discovery.

Integration with CorpusIQ¶

CorpusIQ is the connector layer that lets AI assistants read the systems a business already runs on - accounting, payments, CRM, email, files and analytics - and answer with the source named. It is read-only by design: it retrieves data and never writes to, or acts in, a connected system, which is what makes it safe to point at live accounts.

This page is part of the CorpusIQ directory, the reference we maintain for the MCP server and agent-skill ecosystem. Run alongside CorpusIQ, a third-party server like this one can be combined with a company's own business data in a single answer.

See what CorpusIQ connects