Nook & Nest

I started Nook & Nest because a friend of mine was looking for a rental apartment and wanted to figure out how to furnish it. I thought it would be nice to have a website where you could quickly build your floor plan and move furniture around before buying anything or moving in.

I wanted it to be useful, but also something you would enjoy spending time in. I really like the cozy miniature look of Tiny Glade, so that inspired the feeling I was going for. Rather than a technical-looking floor planner, I wanted decorating to feel more like a relaxing game.

The project grew from there. You can build rooms to real measurements, add floors, walls, doors and windows, and choose furniture, appliances, lighting and decorations. You can change colors and materials, right down to placing books and small decorations on shelves.

Getting the furniture looking right took a lot of back and forth. The early models looked like disconnected blobs, and I wasn’t happy with them. We moved toward original Blender models with more recognizable shapes and cleaner textures. Small things like being able to preview furniture before placing it, zoom in close, and easily find items in the library made a big difference too.

When I heard about the WebMCP Challenge, I wanted to take it a step further: let an AI agent actually furnish the apartment with you. Not just suggest a sofa in a chat, but look at the apartment, browse the available furniture, and put together an arrangement inside the planner.

Through WebMCP, an agent can read the layout and measurements, find furniture, and propose a design. You review the changes first, or you can give it permission to edit directly if you want to let it get creative. You can pause its access and undo an arrangement, so you’re still in control of your space.

I built Nook & Nest with help from Codex, using React, TypeScript, Babylon.js and Blender. A lot of the process was trying something, looking at it, and asking what would make it feel better to use. Adding agent support also taught me that the agent needs clear information about the space and clear boundaries around what it can change.

There’s still more I want to improve, especially comparing different layouts and expanding the furniture styles. But the idea is still the same as when I started: make it easier to picture your future home and have some fun decorating it. Now you can invite an AI agent to help, or see what it comes up with on its own.