CoS

Telegram bot that turns household chatter into booked calendars, done tasks, and a fair split of the mental load

Comment

Inspiration

The invention and widespread adoption of the washing machine was one of history's biggest turning points for reducing gender inequality and enabling women's liberation. Today, one of the most significant contributors to gender inequality is the "mental load" carried within the home, a finding backed by recent research from the World Bank, the Journal of Women & Therapy, and the European Sociological Review.

When I interviewed family and friends about this, I understood the pain points much more clearly. It wasn't the chores themselves that were unevenly split. It was noticing what needed to be done, reminding a partner when necessary, and tracking whether it actually got done. Most of the time, this labor was completely invisible.

So I built a Telegram bot with Strands Agents and started testing it in my own household.

What it does

  • Capture — mention a task naturally ("we need to book Mia's dentist appointment before September") and CoS logs it automatically, no commands needed.
  • Assign & remind — it asks who's handling it, then follows up before the deadline so nothing quietly falls through.
  • Close the loop — say it's done in plain language ("I bought eggs, salt and biscuits") and CoS marks the matching tasks complete.
  • Go beyond tracking — ask it to look things up (store hours), book real dates on your calendar (dentist appointments), or compare products to buy (dishwasher detergent, with prices, ratings, and links).
  • Make the imbalance visible — a weekly summary shows who captured, owned, and completed tasks, turning an invisible load into a number both partners can see.

How I built it

I did the development with Claude Code. Since Claude has strong, readily accessible knowledge of Strands Agents, the coding side moved smoothly most of the time. The core loop (capture a task from chat → assign it → remind → close it out) came together first, and I layered on integrations from there: calendar booking, web search, and product lookup.

Challenges we ran into

Security was the biggest challenge, given my background isn't in software development. I had to actively research potential risks rather than relying on intuition, and treat code review as a deliberate, ongoing practice rather than a one-time check

Accomplishments that I'm proud of

  • Going from an idea to a working agent that my own household actually uses day-to-day.
  • Getting the core loop (capture → assign → remind → close) to feel natural enough that people just talk normally, instead of learning bot commands.
  • Shipping real integrations beyond a to-do list (web search, calendar booking and product search) that let CoS actually do things, not just track them.
  • Building something that made an invisible, hard-to-quantify problem (the mental load) visible with actual numbers.

What we learned

Technically, I picked up a lot along the way such as agent best practices, AWS products, and working with various external APIs.

Beyond the technical side, I re-confirmed just how much I personally need something like this. Imagining a household agent beforehand and actually living with one turned out to be very different experiences. Even just walking from the living room to the kitchen, so many things come to mind that need to get done — and afterward, they quietly become a burden. CoS helped a lot even with MVP scope.

What's next for CoS

I plan to keep developing the agent. The core integrations are in place now, but there's a lot of room to grow. I want to share it with my friends & family that I talked during the process. So, first thing will be making CoS multi-tenant.

Built With

Share this project:

Updates

Submission history