[Feature Request] [Helm Chart] deployment support for the Superset MCP Server component

[dafa-wiratama]

Problem Description

With Apache Superset now natively supporting the Model Context Protocol (MCP) via the superset mcp run daemon, engineering teams are beginning to expose Superset's semantic layer, datasets, and execution capabilities directly to LLM-based tools, AI assistants, and IDE integrations (like Claude or Cursor).

Currently, the official Helm chart provides native, first-class deployment blocks for standalone subsystems like supersetNode, supersetWorker, supersetCeleryFlower, and supersetWebsockets. However, there is no native deployment block for the MCP server. To deploy it as an isolated background daemon in a Kubernetes cluster, users currently have to manually extend core components or maintain external wrapper charts.

Proposed Solution

I would like to contribute a native, first-class implementation of the MCP server daemon to the core Helm chart, following the design patterns used for components like supersetWebsockets or supersetCeleryFlower.

This implementation includes:

1. A new top-level supersetMcp configuration block in values.yaml.
2. A dedicated templates/deployment-mcp.yaml and templates/service-mcp.yaml leveraging global image, environment, and volume configurations.
3. Seamless integration into the existing path routing in templates/ingress.yaml to allow path-based routing sharing the root domain (e.g., routing traffic targeting /mcp cleanly to the daemon).

Proposed values.yaml structure addition:

supersetMcp:
  # -- Enables a native Model Context Protocol (MCP) server deployment
  enabled: false
  replicaCount: 1
  # -- Sets the pod disruption budget for supersetMcp pods
  podDisruptionBudget:
    enabled: false
    minAvailable: 1
    maxUnavailable: 1
  # -- Native execution command for the MCP background daemon
  command:
    - "/bin/sh"
    - "-c"
    - ". {{ .Values.configMountPath }}/superset_bootstrap.sh; superset mcp run --host 0.0.0.0 --port 5008"
  service:
    type: ClusterIP
    annotations: {}
    loadBalancerIP: ~
    port: 5008
    nodePort:
      http: nil
  ingress:
    path: /mcp
    pathType: ImplementationSpecific
  resources: {}
  extraContainers: []
  deploymentAnnotations: {}
  affinity: {}
  topologySpreadConstraints: []
  podAnnotations: {}
  podLabels: {}
  strategy: {}
  podSecurityContext: {}
  containerSecurityContext: {}
  startupProbe:
    httpGet:
      path: /health
      port: mcp
    initialDelaySeconds: 5
    timeoutSeconds: 1
    failureThreshold: 60
    periodSeconds: 5
    successThreshold: 1
  livenessProbe:
    httpGet:
      path: /health
      port: mcp
    initialDelaySeconds: 5
    timeoutSeconds: 1
    failureThreshold: 3
    periodSeconds: 5
    successThreshold: 1
  readinessProbe:
    httpGet:
      path: /health
      port: mcp
    initialDelaySeconds: 5
    timeoutSeconds: 1
    failureThreshold: 3
    periodSeconds: 5
    successThreshold: 1
  priorityClassName: ~