How It Works
The Problem
Household work is invisible. You can't delegate what you can't name. AI agents can't help with tasks that aren't structured, shareable, or even acknowledged. Most of this work lives in one person's head — and that's where it stays until something breaks or someone burns out.
The Approach
Dwellsmith treats household work as a data problem. We give it structure so humans and AI agents can actually work with it:
- → Capture invisible work as structured data across four domains: tasks, relationships, helpers, and home maintenance
- → Share visibility across household members so no one person carries it alone
- → Connect to AI agents through an MCP server with tools to read, create, and complete household work
- → Build toward real household automation — not hype, but structured data that makes delegation possible
What's Built Today
Dwellsmith isn't a concept — it's a working app with a real MCP server. Here's what exists right now:
📋 Tasks
Recurring household tasks with flexible scheduling — daily, weekly, monthly, or custom intervals. Assign to household members. Track what's overdue, due today, and coming up.
👋 Relationships
Track the people you want to stay in touch with. Log calls, visits, and video chats. See who you haven't reached out to in a while.
🤝 Helpers
Household helpers — cleaners, dog walkers, landscapers. Log visits and payments. Keep track of who came and when.
🏠 Home Maintenance
Recurring maintenance items on a schedule. HVAC filters, gutter cleaning, seasonal tasks. See what's overdue and what's coming up.
🔧 MCP Server
13 tools that let Claude (or any MCP-compatible agent) manage your household through conversation. Create tasks, log interactions, complete maintenance — all in plain English.
Real Example: Thanksgiving 2025
Using my cooking-with-claude system, I planned a 7-dish Thanksgiving meal. Here's what making invisible work visible actually looks like:
4-Day Prep Schedule
Breaking one complex meal into structured tasks across 4 days. Hour-by-hour timeline. Storage locations for each ingredient. Exact prep vs. assembly division. That's 50+ invisible micro-decisions made visible and shareable.
This is what household AI needs: Structured workflows, explicit timing, ingredient locations, task dependencies. You can't delegate to an AI agent what you can't articulate to a human collaborator.
Where This Is Going
We're building the semantic layer between human intention and machine execution inside the household. Today that means structured data and an MCP server. Tomorrow it means AI agents that can genuinely manage household work — not just chat about it, but actually do it. That starts with making the invisible visible.