HUB
Sign inSign up

talkopsai/helm-mcp-server

By talkopsai

•Updated 3 days ago

An MCP server that gives AI assistants full control over Helm - from chart discovery to deployments.

Image
Machine learning & AI
Developer tools
0

712

talkopsai/helm-mcp-server repository overview

⁠Helm MCP Server

Give your AI assistant safe, structured access to Helm v3 and Kubernetes. Search charts, validate values, preview diffs, install or upgrade releases, and roll back broken deployments—with secrets redacted automatically.

⁠Quick Reference

⁠What You Get

  • 13 MCP tools (7 read-only) built for AI agents and MCP clients
  • 10 MCP resources under helm:// for structured context discovery
  • 4 Agent Skills bundled in /app/skills (browse on GitHub⁠)
  • Container ready: Includes helm v3 and kubectl inside a minimal non-root image (linux/amd64, linux/arm64).
  • Safe by default: starts in read-only mode until you explicitly set MCP_ALLOW_WRITE=true.

⁠Quick Start (MCP Client over stdio)

Use this configuration in Claude Desktop, Cursor, Windsurf, or VS Code (mcp.json / claude_desktop_config.json):

{
  "mcpServers": {
    "helm": {
      "command": "docker",
      "args": [
        "run",
        "-i",
        "--rm",
        "-e",
        "MCP_TRANSPORT=stdio",
        "-v",
        "/path/to/your/kubeconfig:/app/.kube/config:ro",
        "talkopsai/helm-mcp-server:latest"
      ]
    }
  }
}

Using Claude Code? Register it in one command:

claude mcp add helm -- docker run -i --rm -e MCP_TRANSPORT=stdio -v ~/.kube/config:/app/.kube/config:ro talkopsai/helm-mcp-server:latest

Tip: If your kubeconfig references external client certificate files (for example with minikube or kind), mount the full ~/.kube directory (-v /path/to/.kube:/app/.kube:ro) so those files are reachable inside the container.

Restart your MCP client and try asking:

  • "Install the latest Bitnami PostgreSQL chart into the database namespace."
  • "Show me what would change if I upgrade my-app to chart version 2.1.0."
  • "My redis release is failing. What's wrong and how do I roll it back?"

⁠Running Standalone (Streamable HTTP)

By default (MCP_TRANSPORT=http), the container starts an MCP Streamable HTTP server on port 8765 so multiple clients can share a single instance:

docker run --rm -p 8765:8765 \
  -v ~/.kube/config:/app/.kube/config:ro \
  talkopsai/helm-mcp-server:latest

Point your MCP client at http://localhost:8765/mcp:

{
  "mcpServers": {
    "helm": {
      "url": "http://localhost:8765/mcp"
    }
  }
}

⁠Allowing Changes (Write Mode)

Out of the box, the container runs in read-only mode (MCP_ALLOW_WRITE=false). To allow mutations (installing, upgrading, rolling back, or uninstalling releases and adding chart repositories), pass -e MCP_ALLOW_WRITE=true to docker run. Even with writes enabled, changes support dry-run previews and destructive operations require explicit confirmation.


⁠Environment Variables

VariableContainer DefaultDescription
KUBECONFIG/app/.kube/configPath to mounted kubeconfig inside the container
MCP_ALLOW_WRITEfalseSet to true to enable install, upgrade, rollback, and uninstall tools
MCP_TOOLSETSallFilter tools: read, mutate, or all
HELM_TIMEOUT300Timeout in seconds for Helm CLI operations
MCP_TRANSPORThttphttp (standalone server) or stdio (MCP client subprocess)
MCP_PORT8765HTTP listen port inside the container

⁠Available Tools (13)

ToolAccessDescription
helm_ensure_repositoryWriteAdd and index a Helm chart repository if not already configured.
helm_get_release_statusReadGet concise current status, revision, resource counts, and resource URIs for a Helm release.
helm_install_chartWrite (destructive)Install a new Helm release onto the active Kubernetes cluster context.
helm_list_releasesReadList Helm releases in the active cluster with optional namespace/status filtering and pagination.
helm_monitor_deploymentReadEvaluate Helm release status, Deployment/StatefulSet rollouts, and container failure states.
helm_plan_releaseReadCompute a pre-deployment plan for a Helm release (install preview if new; resource/image/replica diff if existing).
helm_render_manifestsReadRender Kubernetes manifests from a Helm chart offline with automatic Secret data redaction.
helm_rollback_releaseWrite (destructive)Roll back a Helm release to a previous revision (or simulate with dry_run=True).
helm_search_chartsReadSearch configured Helm chart repositories by keyword.
helm_uninstall_releaseWrite (destructive)Uninstall a Helm release and delete its managed Kubernetes resources.
helm_upgrade_releaseWrite (destructive)Upgrade an existing Helm release to a new chart version or values configuration.
kubernetes_check_prerequisitesReadVerify whether the active cluster meets minimum Kubernetes version and API resource/CRD requirements.
kubernetes_set_contextWriteSwitch the active Kubernetes cluster context for all subsequent Helm and Kubernetes calls in this session.

⁠Bundled Agent Skills (4)

This image also ships step-by-step operational playbooks (SKILL.md) under /app/skills:

To copy the skills onto your workstation for Claude Code or Cursor:

docker run --rm --entrypoint tar talkopsai/helm-mcp-server:latest -C /app/skills -cf - . | tar -xf - -C ~/.claude/skills/

⁠Support & Community

Tag summary

Content type

Image

Digest

sha256:07690b576…

Size

118 MB

Last updated

3 days ago

docker pull talkopsai/helm-mcp-server