convez.ai
AgenteAutomaçãoAvançado
aaaa-zhen

Claude Home Agent

aaaa-zhen/claude-home-agent

Controle sua vida via WeChat com um assistente AI persistente

Ver no GitHub
BomPluginScore · de 0 a 100

Agente de IA que funciona como mordomo pessoal via WeChat, integrando controle de casa inteligente (Home Assistant), navegação, compras, viagens e muito mais. Construído com Claude Code, executa automações proativas 24/7 e mantém contexto entre sessões.

Instalação

$npx convez-ai aaaa-zhen/claude-home-agent

1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.

Métricas

Estrelas
11
Forks
0
Acompanhando
0
Saúde
42%
Último commit
há 1 mês
Versão
—
Issues abertas
0
Contribuidores
1

Casos de uso

  • Automação proativa de clima e iluminação baseada em contexto
  • Controle de casa inteligente (AC, luzes, cortinas) por chat
  • Navegação com geofencing e lembretes contextuais de localização
  • Agendamento de rides, consulta de voos e rastreamento de encomendas
  • Assistente multimodal que processa fotos (recibos, menus, documentos)

Veredito

Projeto sofisticado que demonstra real-world aplicação de agents com Claude Code em ecosistema chinês. Arquitetura bem pensada (WeChat ↔ Claude Code ↔ Home Assistant), mas requer setup complexo e infraestrutura específica (servidor Linux, HA instance). Excelente referência de automação residencial proativa.

README

A tradução para o português deste README ainda está na fila.

Claude Home Agent

Talk to your AI butler over WeChat — control smart home, hail rides, order food, navigate, check tickets, and more.

Built on Claude Code + Home Assistant. One chat handles everything.

https://github.com/user-attachments/assets/0f30d916-d256-4519-9beb-451c7c89673f

📱 Want a generative-UI iOS frontend (Siri-style cards) for the same agent? See siri-agent-app.


Highlights

The core idea: a single CLAUDE.md turns Claude Code into your personal butler. No heavy code — it's mostly prompt engineering. The agent doesn't just reply, it acts on its own and remembers across sessions.

AreaWhat it does
🤖 Proactive automationWatches room temp while you sleep & auto-runs AC; detects arrive/leave home; reminds you of devices left on after you leave; 24/7 background process
🏠 Smart homeAC (on/off/temp/mode, multi-unit), lights (per-room), HomePod music, curtains, live status queries
🚗 NavigationRoutes from your live GPS; searches nearby restaurants/cafes/malls with one-tap Amap deep links; driving / walking / cycling
📍 Geofence reminders"Remind me to buy milk at the supermarket" — checks GPS every minute, pings WeChat on arrival
🚕 Ride hailingSearch destination, quote, book a DiDi, track the driver
✈️ TravelFlights (price/time), high-speed rail (live seats via 12306), weather
📦 ShippingCompare courier quotes (SF/ZTO/YTO/Yunda), book SF pickup
📹 Media & filesSend a link → auto-download (X / YouTube / Douyin); built-in FFmpeg for clip/convert/compress
⏰ Reminders"Remind me at 8am tomorrow" — writes to crontab, self-clears after firing
📖 Daily English readingAuto-fetches articles (Reddit/BBC/Guardian), AI-generates B2 reading cards
🎨 Image generation"Draw me a…" via gpt-image, sent back to WeChat
🧠 Persistent memoryRemembers preferences, devices, history; builds a user profile; learns from corrections
📷 Image recognitionSend a photo (receipt/menu/doc/product) — native multimodal understanding
🔀 Model switching"Switch to Opus / Sonnet" in one message

How it works

WeChat message → weixin-acp → Claude Code CLI → Home Assistant / various APIs → reply to WeChat

Quick start

Prerequisites: a Linux server (1GB VPS is fine), Node.js 20+, Python 3.12+, a Home Assistant instance, Claude Code CLI (Anthropic API key), and the WeChat PC client.

git clone https://github.com/aaaa-zhen/claude-home-agent.git
cd claude-home-agent

cp .env.example .env          # fill your keys
cp -r memory-templates/ memory/   # init memory
cp CLAUDE.md.example CLAUDE.md     # customize your agent

python3 -m venv venv && source venv/bin/activate
pip install requests python-dotenv
npm install -g weixin-acp

./start.sh                    # scan the QR to log in WeChat, then message yourself

Roadmap

  • Cars — Lynk&Co / BYD / Tesla via HA: check battery & location, pre-cool/heat, push nav to the car
  • More platforms (MCP/CLI) — Meituan/Ele.me food, JD/Taobao shopping, Apple Health, calendar sync, bills
  • Smart home — Zigbee/Matter devices, smart locks, security cams with motion push, energy monitoring
  • Multi-user — per-member WeChat binding, role-based access, elder/kid modes
  • Voice — ESP32-S3 / Raspberry Pi wake-word entry with local ASR

License

MIT


用微信跟 AI 管家对话,控制智能家居、叫车、点餐、导航、查票…… 基于 Claude Code + Home Assistant,一条微信搞定所有事。

想要同一个 agent 的生成式 UI iOS 前端(Siri 风格卡片)?见 siri-agent-app。

核心思路:用一个 CLAUDE.md 把 Claude Code 变成你的私人管家,几乎不写代码、全靠 prompt engineering。Agent 不只被动回复,还会主动出击、跨会话记忆。

能力说明
🤖 主动自动化睡觉时守护室温自动开空调;到家/离家感知;离家忘关设备提醒;7×24 后台运行
🏠 智能家居空调(开关/调温/模式/多台)、灯光(分区)、HomePod 音乐、窗帘、状态查询
🚗 位置感知导航以当前 GPS 为起点规划;搜周边餐厅/咖啡/商场,带高德一键导航;驾车/步行/骑行
📍 地理围栏提醒"到超市提醒我买牛奶" —— 每分钟查 GPS,到达发微信
🚕 叫车搜目的地、查报价、下单滴滴、实时查司机位置
✈️ 出行查询航班(价格/时刻)、高铁(12306 实时余票)、天气
📦 快递多家比价(顺丰/中通/圆通/韵达),顺丰下单寄件
📹 视频/文件发链接自动下载(X/YouTube/抖音);内置 FFmpeg 剪辑转换压缩
⏰ 定时提醒"明早 8 点提醒我开会" —— 写 crontab,触发后自动清除
📖 每日英语阅读自动抓 Reddit/BBC/Guardian,AI 生成 B2 阅读卡片
🎨 AI 图片生成"帮我画一张……" 直接发回微信
🧠 持久记忆跨会话记住偏好/设备/历史,积累用户画像,从纠正中学习
📷 图片识别发图(快递单/菜单/文件/商品)原生多模态识别
🔀 模型切换"切 Opus / 切 Sonnet" 一句话搞定

工作原理:微信消息 → weixin-acp → Claude Code CLI → Home Assistant / 各类 API → 回复微信

快速开始:git clone 后 cp .env.example .env(填密钥)→ cp -r memory-templates/ memory/ → cp CLAUDE.md.example CLAUDE.md → 装依赖 → ./start.sh 扫码登录微信。

Ver o original no GitHub

Bom para

Relacionados