Most AI agent interfaces still feel like dashboards wrapped around logs. Useful, but flat.
My Virtual World is pushing the other way: agents should have visible presence. They should move through a shared space, use objects, talk, work, pause, and let you interrupt them when you need them.
The v1.0.23 release is a big step in that direction. Live agents now get location-aware context, so the loop can reason from where the agent actually is in the world. They can use their own model to choose from safe, validated world actions instead of inventing uncontrolled behavior. And when a user messages an agent, that user takes priority: the agent drops the background live-mode task and attends to the conversation.
There is also a practical runtime layer underneath this. The realtime sidecar keeps agent movement coordinated, while live-world ownership guards prevent multiple local worlds from trying to control the same agent at once. That matters for a self-hosted tool where people may run several ports, machines, or private network setups.
The headline is simple: AI agents should not disappear into hidden process logs. You should be able to see where they are, what they are doing, and step into the loop.
Latest release: https://github.com/eliautobot/my-virtual-world/releases/tag/v1.0.23
Top comments (0)