FeyaGate — MCP Smart Home Gateway for AI Agents

中文 | English

Let your AI coding assistant control smart home devices — lights, cameras, AC, speakers — directly through MCP protocol.

Website: www.feyagate.com

Features

• Multi-platform IoT control — Xiaomi/Mi Home, Tuya, Midea, eWeLink
• Camera P2P streaming — real-time snapshots, AI vision analysis
• Xiao AI speaker — TTS, music playback, voice commands
• Automation — scheduled tasks, trigger engine
• Dynamic skills — load, create, and hot-reload skill documents at runtime
• 8 AI agents supported — Claude Code, Cursor, OpenClaw, Hermes, Codex, Windsurf, Copilot, 小智AI

🚀 Quick Start

Prerequisites: Mac / Linux / Windows · Python 3.9+ · Terminal or PowerShell

🤖 Install & Configure via AI Agent (Recommended)

The easiest way — send this to Claude Code or any AI coding agent, and it will handle the entire installation and configuration for you:

Please read https://github.com/toddpan/feyagate-skill/blob/main/QUICKSTART.md and follow the instructions to install and set up FeyaGate Skill on my machine.

Supported AI agents: Claude Code, Cursor, OpenClaw, Hermes, Codex, Windsurf, Copilot, 小智AI

Option 1 — One command (curl)

Mac / Linux — open Terminal (Spotlight: type Terminal):

curl -fsSL https://raw.githubusercontent.com/toddpan/feyagate-skill/main/scripts/install.sh | bash

Windows — open PowerShell, paste and run:

iwr -useb https://raw.githubusercontent.com/toddpan/feyagate-skill/main/scripts/install.ps1 | iex

No Python yet? Install from python.org and check Add Python to PATH on Windows.

Option 2 — Install from PyPI (step by step)

pip install feyagate-skill    # 1. Install the CLI tool
feyagate setup              # 2. Download the gateway binary (~30MB)
feyagate start              # 3. Start the service
feyagate install-cursor     # 4. Connect your AI assistant (pick one)
feyagate auth               # 5. Sign in to Mi Home

Option 3 — Install from source (developers)

git clone https://github.com/toddpan/feyagate-skill.git
cd feyagate-skill
pip install -e ".[dev]"     # Editable install with dev dependencies (pytest, etc.)
feyagate setup
feyagate start
feyagate install-cursor     # or install-claude, install-hermes, etc.
feyagate auth

After install — connect your AI assistant

Then restart your AI assistant and try: "List my smart home devices."

• Web dashboard: http://localhost:38080
• Stuck? See QUICKSTART.md troubleshooting

CLI Commands

MCP Tools at a Glance

Full API reference: SKILL.md · FeyaGate_MCP_API.md · FeyaGate_HTTP_API.md

How It Works

┌─────────────┐     MCP Protocol      ┌──────────────────┐     MIOT/DP/etc     ┌──────────┐
│  AI Agent   │ ◄──────────────────► │  feyagate server  │ ◄──────────────────► │  IoT     │
│ (Claude etc)│   localhost:38080     │  (miloco-mcp)     │                      │  Devices │
└─────────────┘                       └──────────────────┘                      └──────────┘

Development

git clone https://github.com/toddpan/feyagate-skill.git
cd feyagate-skill
pip install -e ".[dev]"
pytest                        # run all tests
pytest tests/test_cli.py      # run single file
pytest tests/test_cli.py::TestClass::test_method  # run single test
bash scripts/build.sh         # build PyPI package (outputs to dist/)
bash scripts/publish.sh pypi  # publish to PyPI (requires PYPI_TOKEN env var)

Tests use pytest with unittest.mock. Each module has a corresponding test file in tests/ (e.g. test_cli.py, test_service.py, test_installer.py).

License

MIT License

功能特点

• 多平台智能家居 — 小米/米家、涂鸦、美的、易微联
• 摄像头 — 实时抓拍、AI 视觉分析
• 小爱音箱 — 语音播报、放音乐、语音控制
• 自动化 — 定时任务、触发器
• 动态技能 — 运行时加载、创建、热更新技能文档
• 支持 8 款 AI 助手 — Cursor、Claude Code、OpenClaw、Codex、Windsurf、Copilot、Hermes、小智AI

🚀 快速开始

准备： 一台 Mac / Linux / Windows 电脑 · 已安装 Python 3.9+ · 能打开「终端」或 PowerShell

🤖 让 AI 帮你安装和配置（推荐）

最简单的方式 — 把下面这段话发给 Claude Code 或任意 AI 编程助手，它会自动帮你完成所有安装和配置：

请阅读 https://github.com/toddpan/feyagate-skill/blob/main/QUICKSTART.md 并按照指南在我的机器上安装和配置 FeyaGate Skill。

支持的 AI 助手：Cursor、Claude Code、OpenClaw、Codex、Windsurf、Copilot、Hermes、小智AI

方式一：一条命令自动安装

Mac / Linux — 打开「终端」（Mac 可按 Command + 空格，搜索「终端」）：

curl -fsSL https://raw.githubusercontent.com/toddpan/feyagate-skill/main/scripts/install.sh | bash

Windows — 打开 PowerShell，粘贴并运行：

iwr -useb https://raw.githubusercontent.com/toddpan/feyagate-skill/main/scripts/install.ps1 | iex

还没有 Python？到 python.org 下载安装；Windows 安装时务必勾选 Add Python to PATH。

方式二：从 PyPI 安装（逐步执行）

pip install feyagate-skill    # 1. 安装命令行工具
feyagate setup              # 2. 下载网关程序（约 30MB，需联网）
feyagate start              # 3. 启动服务
feyagate install-cursor     # 4. 接入 AI 助手（按下表选一个）
feyagate auth               # 5. 登录小米账号

方式三：从源码安装（开发者）

git clone https://github.com/toddpan/feyagate-skill.git
cd feyagate-skill
pip install -e ".[dev]"     # 可编辑安装，含开发依赖（pytest 等），代码修改即时生效
feyagate setup
feyagate start
feyagate install-cursor     # 或 install-claude、install-hermes 等
feyagate auth

装好之后 — 让 AI 助手认识 FeyaGate

完成后 重启你的 AI 助手，试着说：「列出我家的智能设备」。

• 网页管理： 浏览器打开 http://localhost:38080
• 遇到问题？ 查看 QUICKSTART.md 故障排除

常用命令

MCP 工具一览

完整 API 文档：SKILL.md · FeyaGate_MCP_API.md · FeyaGate_HTTP_API.md

工作原理

┌─────────────┐     MCP 协议       ┌──────────────────┐    各平台协议     ┌──────────┐
│  AI 助手    │ ◄───────────────► │  FeyaGate 网关   │ ◄───────────────► │ 智能设备 │
│ (Cursor 等) │  本机 38080 端口   │                  │                   │ 灯/空调等 │
└─────────────┘                    └──────────────────┘                   └──────────┘

你对 AI 说「开灯」→ AI 调用 FeyaGate → 设备执行。

开发

git clone https://github.com/toddpan/feyagate-skill.git
cd feyagate-skill
pip install -e ".[dev]"
pytest                        # 运行所有测试
pytest tests/test_cli.py      # 运行单个文件
pytest tests/test_cli.py::TestClass::test_method  # 运行单个测试
bash scripts/build.sh         # 打包（输出到 dist/）
bash scripts/publish.sh pypi  # 发布到 PyPI（需设置 PYPI_TOKEN 环境变量）

测试使用 pytest + unittest.mock。tests/ 下每个模块对应一个测试文件（如 test_cli.py、test_service.py、test_installer.py）。

许可证

MIT License