{
  "transport": "streamable_http",
  "url": "http://192.168.1.176:8000/mcp",
  "user_key": { "header": "Authorization", "prefix": "Bearer " },
  "headers": {
    "Authorization": "Bearer mcp_NFgXQg4IF9V7rQhSxLBrupt3lAsFXJQy",
    "Host": "localhost:8000"
  },
  "groups": {
    "read": [
      "panel_overview",
      "server_details",
      "server_history",
      "services_overview",
      "service_incidents",
      "events_recent"
    ]
  },
  "instructions": "MCP tools for hard-panel (GHard Monitor) — server monitoring panel. Read-only observer: servers CPU/RAM/disks, services health, NAS shares, events log. Query mapping: 'how are my servers?' → panel_overview (one call: servers+services+shares); 'what with <server>?' → server_details (id from panel_overview); disk trend → server_history (hours: 1/6/24/168); service down? → services_overview then service_incidents; what happened recently → events_recent. Start with panel_overview when unsure. Statuses: server online/offline; service up/degraded/down/pending (mode=maintenance/readonly/draining is a work mode, not an outage). Event severities: info/warning/critical. IDs must come from overview tools — never guess. Translate technical slugs into human language. Read-only: changes are made by the human via the web UI (https://hpanel.gnexus.space)."
}