Inspiration
I wanted AI interaction to feel visible and collaborative. Verdant turns an agent into a gardening companion you can watch work alongside you.
What it does
Verdant is a relaxing 3D garden where people and AI agents can plant, water, harvest, and plan layouts together. It offers 13 WebMCP tools, persistent plant growth, care recommendations, and animated robot helpers.
How we built it
I used React, TypeScript, React Three Fiber, Zustand, and WebMCP. Human clicks and agent tools share the same validated garden actions, while Cloudflare Workers hosts the finished experience.
Challenges we ran into
The hardest part was making agent actions reliable, observable, and safe. I built background jobs, plot locking, aisle routing, progress tracking, and plan previews so multiple robots can work without conflicts.
Accomplishments that we're proud of
The AI feels like part of the garden rather than a hidden automation layer. Every action appears in the scene and garden log, and users can review a proposed layout before approving it.
What we learned
I learned that agent experiences need strong feedback, shared state, and clear approval points. Visualizing an agent's intent makes the interaction easier to understand and trust.
What's next for verdant
I plan to add more plants, seasonal events, larger gardens, and deeper agent-assisted planning.
Log in or sign up for Devpost to join the conversation.