household-agent
Enables AI-powered household management including inventory tracking, restock predictions, meal planning from available ingredients, and baby supply monitoring through natural language commands.
README
Household Agent
English | 🌐 中文
AI-powered household management — let AI handle the small daily decisions so you don't have to.
Track groceries, predict restocking needs, plan meals from what's in your fridge, and monitor baby supplies — all through Claude Code (MCP), CLI, or a local web dashboard.
Quick Start
npm install
npm start # Web dashboard at http://localhost:3333
Usage
Claude Code (MCP)
Add to your Claude Code MCP config (~/.claude/settings.json):
{
"mcpServers": {
"household": {
"command": "node",
"args": ["/path/to/household/skills/mcp-server.js"]
}
}
}
Then just talk naturally:
- "What groceries do we have?" →
inventory_list - "What do we need to buy this week?" →
shopping_list - "How many days of baby formula left?" →
baby_supply_status - "Log 3 diapers used" →
baby_log_event - "What can I cook with what's in the fridge?" →
meal_suggestions
OpenClaw
Import the skills/ folder via OpenClaw Skill Manager, or add MCP config:
command: node
args: [/path/to/household/skills/mcp-server.js]
Skill metadata is defined in skills/SKILL.md.
CLI
node skills/inventory.js add <barcode or name> # Add item (auto-lookup via Open Food Facts)
node skills/inventory.js consume <ID or name> [qty] # Record consumption
node skills/inventory.js list [location] # View inventory
node skills/inventory.js expiring [days] # Expiring items (default 7 days)
node skills/inventory.js status # Dashboard summary
Web Dashboard
http://localhost:3333 — tabs for inventory, restock, cooking, calendar, baby tracking, and settings.
Architecture
household/
├── skills/
│ ├── lib/
│ │ ├── data.js # Shared data layer (atomic JSON I/O, paths, utils)
│ │ ├── inventory-ops.js # Inventory CRUD + restock logic
│ │ ├── baby-ops.js # Baby log operations
│ │ └── meal-ops.js # Meal diary + cooking recommendations
│ ├── mcp-server.js # MCP server (12 tools, stdio transport)
│ ├── server.js # Express web server + API
│ ├── inventory.js # CLI entry point
│ ├── predict.js # Consumption prediction engine
│ └── public/index.html # Single-page web dashboard
├── data/ # JSON data files (gitignored)
└── config/ # Category, preference, and location configs
Three entry points — MCP server, web server, and CLI — share a common data layer (lib/), operating on the same JSON files.
MCP Tools
| Tool | Description |
|---|---|
inventory_list |
List inventory with filters (location, category, status) |
inventory_add |
Add item with optional barcode lookup |
inventory_consume |
Record consumption by ID or name |
inventory_expiring |
Items expiring within N days |
inventory_status |
Summary stats (totals, by location/category) |
restock_recommendations |
Smart restock suggestions based on consumption rate |
meal_suggestions |
What to cook based on available ingredients |
shopping_list |
Combined shopping list (restock + expiring + recipe gaps) |
baby_log_event |
Log baby events (feeding, diaper, sleep, etc.) |
baby_supply_status |
Baby supply levels and days-until-empty predictions |
preferences_get |
Read household preferences |
preferences_update |
Update a preference field |
Key Features
- Consumption prediction — tracks usage patterns to predict when items run out
- Smart restock — auto-generates shopping lists based on consumption velocity
- Recipe gap matching — finds dishes you can almost make (missing just 1 ingredient)
- Baby tracking — feeding, diaper, sleep logging with supply predictions
- Barcode scanning — auto-lookup via Open Food Facts API
- Image recognition — identify items via Claude Vision (web dashboard)
Baby Tracker Import
Import baby life records from Baby Tracker app (.btcp format):
node skills/import-btcp.js <file.btcp> [--dry-run] [--tz=8]
Or upload via web dashboard at Settings → Import.
Data Storage
All data is local JSON files — no cloud, no database server, no account required. Atomic writes (write-tmp-then-rename) prevent corruption.
License
Apache License 2.0 — see LICENSE.
Premium features and add-on services may be offered under separate terms.
推荐服务器
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。
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。
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。
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。
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。
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。
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。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。