Most AI agent tools still feel like a terminal with extra steps.
You ask for work, wait, and then get a wall of text. Maybe logs if you dig. Maybe a status chip if the app is fancy.
I wanted the opposite: a place where agents are visible.
So I released My Virtual World: a self-hosted 3D world where AI agents can live, move between buildings, meet, chat, show activity, and make local agent work easier to inspect.
What it does:
- 3D voxel-style world rendered with Three.js
- Roads, buildings, furnished interiors, outside spaces, agents, and object interactions
- Agent movement, seating, service queues, meetings, chat bubbles, and world actions
- Optional OpenClaw, Hermes, and Codex integrations for live agent presence
- Docker-first local install
- Demo mode from the public repo, with an Early Bird full build for the premium builder
The bigger idea is simple:
If AI work is becoming a team, it should not be invisible.
You should be able to zoom out and see who is active, where work is happening, what state each agent is in, and which workflows need attention.
Website: https://myvirtualworld.ai
GitHub: https://github.com/eliautobot/my-virtual-world
I would love feedback from people building agent tools, local-first apps, self-hosted software, and interfaces that make computers feel alive.
Maybe the future of agent ops is not another dashboard.
Maybe it is a world.
