kubectl MCP Server
MCP server that bridges AI assistants to Kubernetes clusters, exposing 253 tools for pod diagnostics, deployment management, RBAC auditing, Helm chart management, cost optimization, and network diagnostics through natural language conversations.
Score Breakdown
⚙ Agent Friendliness
🔒 Security
Uses kubeconfig — inherits cluster RBAC. Agents with kubectl access have broad cluster control. Use dedicated service account with minimal RBAC. Never give cluster-admin to agent.
⚡ Reliability
Best When
You want to interact with Kubernetes clusters conversationally for debugging, auditing, or management tasks without memorizing kubectl commands.
Avoid When
You need deterministic GitOps workflows, have strict change management policies requiring human review, or operate in air-gapped environments. Use ArgoCD or Flux instead.
Use Cases
- • Debugging crashed pods and analyzing container logs via AI assistant
- • Deploying and scaling applications through natural language commands
- • Auditing RBAC permissions and security configurations across clusters
- • Identifying resource waste and optimizing cluster costs
- • Managing Helm releases and viewing interactive dashboards
- • Multi-cluster management from a single AI interface
Not For
- • Production GitOps pipelines where changes must go through version control
- • Environments where AI-initiated cluster changes are prohibited by policy
- • Air-gapped clusters without MCP client connectivity
Interface
Authentication
Uses existing kubeconfig for cluster access. Enterprise features include OAuth 2.1 authentication, RBAC validation, non-destructive mode, and secret masking. The server inherits whatever Kubernetes RBAC permissions the configured kubeconfig grants.
Pricing
MIT licensed. Available via npm and PyPI. Docker image also available.
Agent Metadata
Known Gotchas
- ⚠ 253 tools is extremely high - may overwhelm agent tool selection and consume significant context window
- ⚠ Write operations (scale, delete, deploy) can cause real cluster damage if non-destructive mode is not enabled
- ⚠ Secret masking must be explicitly enabled to prevent leaking sensitive values to LLM context
- ⚠ Requires kubectl and kubeconfig pre-configured - the MCP server cannot set up cluster access
- ⚠ Python and Node.js installation options have different feature sets (UI dashboards only in Python pip install)
Alternatives
Full Evaluation Report
Detailed scoring breakdown, competitive positioning, security analysis, and improvement recommendations for kubectl MCP Server.
Scores are editorial opinions as of 2026-03-06.