Homey-Wan-Kenobi-MCP

Homey-Wan-Kenobi-MCP

MCP server for controlling Homey Pro smart home systems, offering 60+ tools for device control, flow authoring, automation, monitoring, and network diagnostics.

Category
访问服务器

README

Homey-Wan-Kenobi MCP

CI Docker Pulls License: MIT

MCP server for controlling Homey Pro smart home systems through the Model Context Protocol.

The name nods to Obi-Wan Kenobi, your home's only hope for natural-language control.

60 tools and 3 knowledge prompts for device control, flow authoring, automation, monitoring, troubleshooting, network diagnostics, and self-hosted deployment.

Why Homey-Wan-Kenobi

Athom ships an official Homey MCP Server: a cloud-brokered remote connector you add to a paid Claude or ChatGPT plan and sign into with your Homey account. It covers device status and control, renaming and moving devices, creating, updating and starting Flows and Advanced Flows, and setting Moods.

This server differs in two ways. It is self-hosted, so you run it yourself over stdio or streamable-HTTP with no third-party broker and no paid AI-plan requirement. And it adds operational depth for power users that is not part of the official server's described capabilities.

The table reflects Athom's publicly described capabilities as of June 2026. Athom does not publish a formal tool list, so the right-hand gaps are inferred from their documentation, not stated by Athom.

Capability Official Homey MCP (mcp.athom.com) Homey-Wan-Kenobi
Hosting Cloud-brokered remote connector Self-hosted, stdio or HTTP, no broker
Access Paid Claude/ChatGPT connector plan Any MCP client, no plan requirement
Devices, zones, control, rename, move Yes Yes
Flow and Advanced Flow authoring Yes Yes
Moods Yes Yes
Flow-card schema discovery (build flows from scratch) Not described Yes
Energy: live power and day/week/month/year reports Not described Yes
Insights: historical sensor and meter data Not described Yes
Z-Wave and Zigbee mesh diagnostics with remediations Not described Yes
App memory and usage analysis (removal candidates) Not described Yes
Infrastructure ops: backups, reboot, updates, memory, storage Not described Yes
Raw Homey Web API passthrough Not described Yes
Listed in the public MCP Registry No Yes

The depth, in practice

  • Network diagnostics that the official server does not describe. diagnose_zwave_network and diagnose_zigbee_network read controller health, grade each node by last-seen age (battery-sleep aware), group transmit failures, correlate them with unavailable devices, and return severity-ranked remediations. get_zwave_log exposes the raw network log.
  • Energy and Insights analysis. Live power by zone and device, day/week/month/year energy reports, and timestamped historical sensor and meter data, so the model can answer "what is using power right now" or "how has the bedroom temperature trended this week".
  • App memory and usage analysis. analyze_app_usage cross-references every app against RAM, device counts and flow references to flag safe removal candidates and estimate savings. Read-only and advisory.
  • Build automations from scratch. A flow-card discovery engine (list_flow_cards, get_flow_card) returns each card's full argument schema, and full standard and Advanced Flow CRUD (including delete) lets an AI author and tear down the whole cards graph by UUID.
  • Infrastructure operations from one interface. Backups, reboot, update checks, memory and storage usage, drivers, LED ring, session and location info.
  • A future-proof escape hatch. homey_api_call reaches any Homey Web API endpoint directly, so new firmware features work without waiting for a tool update.

It runs on a hardened, non-root node:22-alpine image whose every build and release is gated on a Trivy scan, and it is published to Docker Hub, GHCR and the public MCP Registry. See the tool reference below for the full list.

When the official server is the better fit

This server is local-first and self-hosted, which is both its strength and its trade-off. The official Homey MCP is the easier choice when you want:

  • Zero setup and managed hosting. Add one URL and sign in with your Homey account. There is no container to run, expose, or maintain, and there is a one-click ChatGPT app.
  • Secure remote access from anywhere. Athom brokers the connection through its cloud, so it works away from home without exposing anything on your LAN. This server's HTTP mode is intended for a trusted local network; remote use is your own VPN or authenticated reverse proxy.
  • Every Homey model. The official endpoint reaches Homey Cloud, Pro, Pro mini and Self-Hosted Server, including cloud-only setups that have no local API. This server targets a Homey reachable over its local API or your Athom token, so it is happiest with a Homey Pro.
  • First-party support. It is maintained by Athom, kept in step with firmware, and officially tested with Claude and the ChatGPT app. This project is independent and best-effort.

The two are complementary. Many people use the official connector for quick remote control and this server for the deep local diagnostics, energy analysis, and infrastructure work. Athom does not publish a formal tool list, so the official server may also include capabilities not reflected above.

Quick Start

git clone https://github.com/Ringosystems/Homey-Wan-Kenobi-MCP.git
cd Homey-Wan-Kenobi-MCP
npm install
npm run build

Authenticate with your Homey:

npx homey login
npx homey select

Add to your MCP client config (Kiro, Claude Desktop, Cline, etc.):

{
  "mcpServers": {
    "homey": {
      "command": "node",
      "args": ["/path/to/Homey-Wan-Kenobi-MCP/dist/index.js"]
    }
  }
}

Authentication

The server supports two authentication methods:

Homey CLI (recommended) runs npx homey login and npx homey select. The server reads the stored OAuth token from ~/.athom-cli/settings.json automatically.

Local API Key is created at my.homey.app. Set environment variables:

export HOMEY_ADDRESS=http://192.168.1.x
export HOMEY_TOKEN=your-api-key

The local API key method is the recommended choice when running self-hosted in a container, since it does not depend on a mounted CLI settings file.

Self-Hosted (Docker / HTTP)

The image is published at ringosystems/homey-wan-kenobi (mirrored to ghcr.io/ringosystems/homey-wan-kenobi), and the server is listed in the MCP Registry as io.github.Ringosystems/homey-wan-kenobi-mcp. It is built on node:22-alpine, runs as a non-root user, and ships no baked-in secrets. See SECURITY.md for the audit details.

It defaults to the stdio transport so an MCP client can launch it directly, and exposes a long-lived streamable-HTTP service (via supergateway at /mcp, health at /healthz) when you set MCP_TRANSPORT=streamable-http.

Run as an MCP client (stdio)

docker run -i --rm \
  -e HOMEY_ADDRESS=http://192.168.1.x \
  -e HOMEY_TOKEN=your-local-api-key \
  ringosystems/homey-wan-kenobi:latest

Run as an HTTP service

docker run -d -p 8000:8000 \
  -e MCP_TRANSPORT=streamable-http \
  -e HOMEY_ADDRESS=http://192.168.1.x \
  -e HOMEY_TOKEN=your-local-api-key \
  --restart unless-stopped \
  ringosystems/homey-wan-kenobi:latest

Docker Compose

Supply your Homey credentials via a .env file next to docker-compose.yml:

cat > .env <<'EOF'
HOMEY_ADDRESS=http://192.168.1.x
HOMEY_TOKEN=your-local-api-key
EOF

# Pull the published image, or add --build to build locally instead
docker compose up -d

The provided docker-compose.yml references the published image (with build: . as a local-build fallback) and applies container hardening (read-only root filesystem, dropped Linux capabilities, no-new-privileges, and memory/PID limits):

services:
  homey-wan-kenobi:
    image: ringosystems/homey-wan-kenobi:latest
    build: .
    container_name: homey-wan-kenobi
    ports:
      - "8000:8000"
    environment:
      MCP_TRANSPORT: streamable-http
      HOMEY_ADDRESS: "${HOMEY_ADDRESS:-http://192.168.1.x}"
      HOMEY_TOKEN: "${HOMEY_TOKEN:-}"
    restart: unless-stopped
    read_only: true
    tmpfs:
      - /tmp
    security_opt:
      - no-new-privileges:true
    cap_drop:
      - ALL
    mem_limit: 256m
    pids_limit: 128
    healthcheck:
      test: ["CMD", "node", "-e", "fetch('http://localhost:8000/healthz').then(r=>process.exit(r.ok?0:1)).catch(()=>process.exit(1))"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 20s

Build locally

docker build -t ringosystems/homey-wan-kenobi .
docker run -d -p 8000:8000 \
  -e MCP_TRANSPORT=streamable-http \
  -e HOMEY_ADDRESS=http://192.168.1.x \
  -e HOMEY_TOKEN=your-local-api-key \
  --restart unless-stopped \
  ringosystems/homey-wan-kenobi

Connecting a client

Point an MCP client at the streamable-HTTP endpoint:

{
  "mcpServers": {
    "homey": {
      "type": "streamable-http",
      "url": "http://your-host:8000/mcp"
    }
  }
}

Check health with curl http://your-host:8000/healthz.

Tools

Devices

Tool Description
list_devices List devices with live capability values, filter by zone or class
get_device Get full device details and all capability values by ID
search_devices Search devices by name, class, or capability
set_device_capability Control a device (onoff, dim, target_temperature, volume_set, etc.)

Device & Zone Management

Tool Description
rename_device Rename a device
move_device_to_zone Move a device to a different zone (room)
create_zone Create a new zone, optionally nested under a parent

Zones

Tool Description
list_zones List all zones (rooms/areas) with their hierarchy

Flows

Tool Description
list_flows List simple and advanced flows with enabled/broken status
trigger_flow Run a flow immediately
set_flow_enabled Enable or disable a flow

Flow Authoring

Tool Description
get_flow Get a standard WHEN/AND/THEN flow definition by ID
create_flow Create a standard flow from trigger/conditions/actions
update_flow Update fields of a standard flow
delete_flow Permanently delete a standard flow
get_advanced_flow Get an Advanced Flow definition (cards, args, connections)
create_advanced_flow Create an Advanced Flow from a cards graph
update_advanced_flow Update an Advanced Flow (replaces the cards graph)
delete_advanced_flow Permanently delete an Advanced Flow

Flow Card Discovery

Tool Description
list_flow_cards List trigger/condition/action cards, filtered by substring
get_flow_card Get one flow card's full definition and argument schema

Logic & Apps

Tool Description
list_variables List logic variables with current values
set_variable Set a logic variable value
list_apps List installed apps with version, status, and origin
restart_app Restart a Homey app
enable_app Enable or disable an app
uninstall_app Uninstall an app and remove its devices

Insights & Energy

Tool Description
list_insights List available insight logs
get_insight_entries Get historical sensor/meter data over a time range
get_energy_live Live power consumption by zone and device
get_energy_report Energy report for day/week/month/year

Weather, Presence & Location

Tool Description
get_weather Current weather at Homey's location
get_weather_hourly Hourly weather forecast
get_presence Home/away and awake/asleep status for all users
set_presence Set your own presence or sleep state
get_location Homey's configured geographic location

Alarms & Moods

Tool Description
list_alarms List all alarms and timers
set_alarm Create or update an alarm
delete_alarm Delete an alarm
list_moods List moods (scenes) per zone
set_mood Activate a mood in a zone

Notifications

Tool Description
list_notifications List the 50 most recent notifications
create_notification Send a notification to the Homey timeline

Network Diagnostics

Tool Description
diagnose_zigbee_network Zigbee mesh health, per-node last-seen, issues and remediations
diagnose_zwave_network Z-Wave health, transmit failures, unavailable nodes, remediations
get_zwave_log Raw recent Z-Wave network log for troubleshooting

App Usage Analysis

Tool Description
analyze_app_usage Cross-reference apps against RAM, devices, and flow references to flag removal candidates (advisory, read-only)

System & Infrastructure

Tool Description
get_system_info System info (version, wifi, hostname, hardware)
list_drivers List all available device drivers
get_backup_status Backup config and last backup time
create_backup Schedule a new backup
get_ledring LED ring screensaver options and current setting
set_ledring Set the LED ring screensaver
get_updates Check for system updates and update settings
get_session Current API session info (user, role, scopes)
reboot_homey Reboot the Homey Pro
get_memory_info Memory usage by app and component
get_storage_info Storage usage breakdown
set_system_name Set the Homey system name

Advanced

Tool Description
homey_api_call Raw escape hatch to call any Homey Web API endpoint directly

Prompts

Built-in knowledge prompts accessible via the MCP prompts API:

Prompt Description
homey_best_practices Zone architecture, device naming, protocol tips, energy management, security
homey_troubleshooting Diagnosing offline devices, Z-Wave/Zigbee issues, flow debugging, performance
homey_flow_patterns Automation patterns, naming conventions, anti-patterns to avoid

Development

npm run dev        # Run with tsx (no build step)
npm run build      # Compile TypeScript
npm run lint       # ESLint
npm run typecheck  # TypeScript strict check
npm start          # Run compiled version

Tech Stack

  • TypeScript + Node.js (>=20)
  • homey-api v3, the official Homey Web API client
  • @modelcontextprotocol/sdk v1.29, MCP server (registerTool with ToolAnnotations)
  • zod for parameter validation
  • supergateway for the self-hosted streamable-HTTP transport

Security

The dependency tree, base image, and container configuration are audited with Trivy and npm audit. The published image runs as a non-root user on node:22-alpine with no baked-in secrets, and the production dependencies ship with 0 high or critical advisories. See SECURITY.md for the full audit, accepted residual risks, and how to reproduce the scans.

License

MIT

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