{
  "affected": [
    {
      "package": {
        "ecosystem": "PyPI",
        "name": "ha-mcp"
      },
      "ranges": [
        {
          "events": [
            {
              "introduced": "0"
            },
            {
              "fixed": "7.10.0"
            }
          ],
          "type": "ECOSYSTEM"
        }
      ]
    }
  ],
  "aliases": [],
  "database_specific": {
    "cwe_ids": [
      "CWE-306",
      "CWE-352"
    ],
    "github_reviewed": true,
    "github_reviewed_at": "2026-07-07T23:41:21Z",
    "nvd_published_at": null,
    "severity": "MODERATE"
  },
  "details": "### Summary\n\nIn add-on mode, the ha-mcp settings UI routes are mounted both under the MCP secret path **and** at the bare root of the published port (`:9583`), so Home Assistant ingress can serve the \"Open Web UI\" button. The root-mounted routes perform no authentication \u2014 no secret, no `Origin` check, no CSRF token \u2014 so any client that can reach `:9583` without the MCP secret can invoke them.\n\n### Affected configurations\n\nHome Assistant **add-on** installations (`host_network: true` with port `9583` published), v7.6.0 and earlier. Docker and standalone installs are **not** affected \u2014 there the settings routes are mounted only under the secret path.\n\nRoot-mounted routes in affected versions: tool visibility (`/api/settings/tools` GET/POST), feature flags (`/api/settings/features` GET/POST), the auto-backup suite (`/api/settings/backups\u2026` incl. restore/delete, and `/api/settings/backup-config`), add-on restart (`/api/settings/restart`), and \u2014 when the opt-in Tool Security Policies feature is enabled \u2014 the approval-policy API (`/api/policy/config` GET/PUT, `/api/policy/approve`, `/api/policy/deny`, \u2026).\n\n### Impact\n\nWithout authentication, a caller that reaches `:9583` \u2014 a peer on the local network, a reverse proxy/tunnel that forwards the bare root path (e.g. a whole-host Cloudflared config), or a CSRF `POST` from a page open in a LAN browser \u2014 can read or change which MCP tools are exposed, toggle feature flags, list/view/restore/delete backups, restart the add-on, and (with Tool Security Policies enabled) read and rewrite the approval policy, disabling the human-approval gate on gated tools.\n\nThere is **no** access to Home Assistant data, entities, or credentials, and no code execution. All effects are confined to the add-on\u0027s own configuration and lifecycle and are recoverable. The primary (same-LAN) vector is within the add-on\u0027s documented trusted-network model; remote reachability requires the operator to have reverse-proxied the bare port.\n\n### Proof of concept\n\nWith the add-on running and reachable on `:9583`, from any host that can reach the port without the secret:\n\n```\nGET  /api/settings/tools                 -\u003e 200   (read tool config, no auth)\nPOST /api/settings/tools  {\"states\":{}}  -\u003e 200   (rewrite tool config, no auth / no CSRF token)\nPOST /api/settings/restart               -\u003e 200   (restart the add-on)\n```\n\nThe MCP endpoint itself remains correctly protected by the secret path.\n\n### Patch\n\nFixed in PR homeassistant-ai/ha-mcp#1508 (merged to `master`): the root-mounted add-on routes are restricted to Home Assistant ingress, which always originates from the Supervisor (`172.30.32.2`); every other caller receives `403`. Direct and remote access continue to use the settings UI under the MCP secret path (`\u2026/\u003csecret\u003e/settings`), so the \"Open Web UI\" button, Cloudflared, and the Webhook Proxy add-on are unaffected.\n\nThe fix will ship in the next stable add-on release. If you\u0027d rather have it now, it is already on the dev channel (add-on dev build ` 7.6.0.dev393` or later) \u2014 optional; there\u0027s no need to switch channels just for this, it is a fairly low risk surface and only exposes the web UI for addon mode only.\n\n### Severity\n\n`CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:L` = 6.4 (Moderate). Confidentiality impact is None \u2014 tool config and backups are not secrets or credentials; integrity and availability impacts are Low \u2014 configuration changes and an add-on restart are recoverable.\n\n### Credit\n\nReported by @bharat.",
  "id": "GHSA-q855-8rh5-jfgq",
  "modified": "2026-07-07T23:41:21Z",
  "published": "2026-07-07T23:41:21Z",
  "references": [
    {
      "type": "WEB",
      "url": "https://github.com/homeassistant-ai/ha-mcp/security/advisories/GHSA-q855-8rh5-jfgq"
    },
    {
      "type": "WEB",
      "url": "https://github.com/homeassistant-ai/ha-mcp/pull/1508"
    },
    {
      "type": "WEB",
      "url": "https://github.com/homeassistant-ai/ha-mcp/commit/9f5b085ad4a7b38b067c9da0dc5b45462c4d796e"
    },
    {
      "type": "PACKAGE",
      "url": "https://github.com/homeassistant-ai/ha-mcp"
    }
  ],
  "schema_version": "1.4.0",
  "severity": [
    {
      "score": "CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:L/A:L",
      "type": "CVSS_V3"
    }
  ],
  "summary": "ha-mcp: Add-on settings and policy routes are reachable without authentication at the bare root path"
}