Sardrobe

Inspiration

Capsule wardrobes are ideal for decluttering and making the most out of your clothes. The digital capsule wardrobe apps my wife tried embodied the opposite of the concept; they all felt like shopping apps, with ads and upsells.

What it does

To address these issues, we created Sardrobe. You can snap pictures of your clothes, automatically remove the background in the browser, visually assemble outfits from them, and create packing lists for trips. There’s no social feed, ads, subscription model, or pressure to buy more clothes.

The app also includes WebMCP tools so that you can offload any step to whichever capable agent you have a subscription for, like Claude or Codex. You can even have the agent act as your own personal stylist.

How we built it

We used a full-stack architecture for this:

• Astro and React
• Hono and Cloudflare Workers
• Cloudflare D1 and R2
• Better Auth and Google OAuth
• Drizzle ORM
• Base UI
• ONNX Runtime Web
• Transformers.js
• Vercel AI SDK
• WebMCP

Challenges we ran into

The main challenges were feeding images to ChatGPT via WebMCP, building a full-stack app with Astro, and getting ChatGPT to work consistently and well with UI/UX design.

Accomplishments and lessons learned

Divide and conquer helps a great deal with large applications like these. I'm also proud of the fact that I've built something my wife and I can use daily with our existing AI subscriptions. Per-user software like this feels very personal and rewarding.

What's next for Sardrobe

Avatar outfit previews, on-device planning AI for drafts, and see if I can reduce the number of WebMCP tools while still achieving the same objectives.

Built With

• astro
• chatgpt
• cloudflare
• hono
• react
• storybook
• vercel
• webmcp