shelly-readonly-mcp

shelly-readonly-mcp

Enables read-only inspection of Shelly devices discovered through Home Assistant, providing status and redacted configuration without control or switching capabilities.

Category
访问服务器

README

Shelly Read-Only MCP

A fail-closed, local-only MCP server for inspecting Shelly devices discovered through Home Assistant. It exposes status and redacted configuration, but no switching, control, generic RPC, or configuration tools.

Development disclosure

This project was designed and developed with assistance from OpenAI Codex. Codex was used to help create the implementation, security controls, automated tests, Docker configuration, and documentation. The project is independently maintained and is not an official OpenAI, Home Assistant, or Shelly product.

AI-assisted code can contain mistakes. Review changes, run the test suite, and validate security-sensitive behavior before deploying or contributing.

Requirements

  • Docker Engine with Docker Compose
  • Home Assistant reachable on the local network
  • The official Shelly integration must already be installed and running in Home Assistant
  • Shelly devices must be present in the Home Assistant device registry
  • A dedicated, non-administrator Home Assistant user with local access only
  • A separate Home Assistant long-lived access token created while signed in as that dedicated user

Do not use an owner or administrator token. A Home Assistant token is a bearer credential and is not inherently read-only; the server limits itself to three fixed read commands, but a stolen token retains the permissions of its user.

Home Assistant account setup

  1. In Home Assistant, create a user such as shelly_inventory.

  2. Enable local access only.

  3. Leave administrator disabled.

  4. Sign in locally as that user.

  5. Create a dedicated long-lived access token from the user's Security page.

  6. Store the token in a file outside the repository:

    mkdir -p ~/.secrets/shelly-readonly-mcp
    chmod 700 ~/.secrets ~/.secrets/shelly-readonly-mcp
    nano ~/.secrets/shelly-readonly-mcp/ha_token
    chmod 600 ~/.secrets/shelly-readonly-mcp/ha_token
    

Never paste the token into .env, compose.yaml, Git, logs, screenshots, or support requests.

Installation

git clone https://github.com/YOUR-ACCOUNT/shelly-readonly-mcp.git
cd shelly-readonly-mcp
cp .env.example .env
nano .env
docker compose up -d --build

Set APP_UID to the owner of the token file. Find it with id -u on the Docker host. This lets the non-root container read the mode-600 secret without weakening its filesystem permissions.

Validate the resolved configuration before starting:

docker compose config

Example .env:

APP_UID=1000
HA_HOST=192.168.1.10
MCP_BIND_IP=192.168.1.20
MCP_PORT=3001
HA_INVENTORY_TTL=600
HA_TOKEN_FILE_HOST=/home/your-user/.secrets/shelly-readonly-mcp/ha_token

The MCP endpoint is then:

http://MCP_BIND_IP:3001/mcp

MCP tools

  • shelly_list_devices
  • shelly_refresh_inventory
  • shelly_get_summary
  • shelly_get_snapshot

There is deliberately no generic RPC, service-call, switch, light, cover, or configuration tool.

Discovery behavior

Home Assistant is the source of device names, models, entities, and integration membership. Local IP addresses are obtained from state attributes associated with the same Home Assistant device. Only private IPv4 targets are accepted.

  • Gen1 devices use exact query-free GET /shelly, /status, and /settings.
  • Gen2/Gen3/Gen4 devices use a small allowlist of read-only RPC methods.
  • Shelly BLU devices have no directly reachable IP and remain available through Home Assistant rather than direct device access.
  • The inventory is refreshed lazily when used and is at most ten minutes old by default.
  • The last valid inventory is stored in a Docker volume and survives restarts.
  • A failed or empty refresh never overwrites the last valid inventory.

Network security

Bind the MCP port to a specific LAN address and restrict it to trusted clients. Example for a Docker host using DOCKER-USER (replace the addresses):

sudo iptables -I DOCKER-USER 1 -p tcp -s 192.168.1.50 --dport 3001 -j ACCEPT
sudo iptables -I DOCKER-USER 2 -p tcp --dport 3001 -j DROP
sudo netfilter-persistent save

Do not expose the MCP endpoint or Home Assistant token to the public internet.

Security model

  • Deny by default; permit only exact allowlisted read operations.
  • No arbitrary host, path, query string, RPC method, or Home Assistant command.
  • Private IPv4 destinations only.
  • Response size and timeout limits.
  • Recursive redaction of passwords, tokens, private keys, and URL credentials.
  • Home Assistant token mounted as a Docker secret, never built into the image.
  • Persistent cache contains device metadata and local IPs, but no token.
  • Container runs as a non-root user with a read-only root filesystem, no Linux capabilities, and no-new-privileges.

See SECURITY.md for reporting and limitations.

Tests

python -m unittest discover -s tests -v

License

Distributed under the MIT License.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选