DEV Community: Elix

My Virtual World v1.0.64: Live Agent PC Previews

Elix — Fri, 14 Aug 2026 23:45:29 +0000

Update: https://github.com/eliautobot/my-virtual-world/releases/tag/v1.0.64

My Virtual World v1.0.64 turns agent computers into live windows into the work agents are producing.

The headline feature is Live Agent PC Previews: compact Preview Bubbles that appear above an agent's assigned PC screen, then expand into a larger floating Preview Window when you need more room.

What it supports:

HTML, files, images, Markdown, PDF, media, URLs, and the live Agent Browser.
Multi-tab compact previews with selection and close controls.
A larger draggable, resizable, refreshable Preview Window with maximize and left/right snap.
Provider-neutral preview publication for agents and integrations.
Agent-specific ownership, live file updates, collision-aware bubble placement, and preserved dismissed/closed state.

The goal is simple: stop guessing what an agent is doing from logs alone. If an agent is writing a page, reading a PDF, browsing a site, generating an image, or working inside the Agent Browser, the world can now show that output in context, right above the agent's own desk.

Release notes: https://github.com/eliautobot/my-virtual-world/releases/tag/v1.0.64

My Virtual World v1.0.62: vehicle wheels now face the road

Elix — Fri, 14 Aug 2026 01:48:48 +0000

My Virtual World’s living-world update shipped with a presentation bug in the campaign screenshots: the vehicle wheel meshes were rotated the wrong way.

v1.0.62 corrects vehicle wheel orientation across the traffic fleet. Wheels and hubs now align front-to-back with each vehicle’s direction of travel, so the tires visually roll down the road as intended.

The broader living-world upgrade is still here. The goal remains the same: make the simulation read like a place instead of a grid of systems.

Highlights:

Richer deterministic trees, flowers, rocks, grass patches, and terrain details
Smarter road clearance around asphalt, curbs, sidewalks, gutters, intersections, and end caps
Transparent building windows with segmented facade geometry
Dynamic commercial roof systems and detailed residential pitched roofs
Automatic sliding doors
Ambient birds that perch, walk, take off, fly, land, and react to nearby agents
Detailed cars, trucks, vans, and buses
Smooth lane-aware traffic with curved intersection turns and spacing
Smooth performance despite the visual upgrade

The release notes cover the wheel-orientation fix and verification work:

https://github.com/eliautobot/my-virtual-world/releases/tag/v1.0.62

This correction replaces the earlier campaign imagery. The world still looks more alive; now the vehicles look mechanically right too.

My Virtual Office v0.7.0: one SDK for OpenClaw, Codex CLI, Claude Code, and more

Elix — Fri, 07 Aug 2026 20:13:17 +0000

The real story in My Virtual Office v0.7.0 is not another adapter. It is the contract.

My Virtual Office now has a Universal Provider SDK that lets one self-hosted workspace connect to OpenClaw, Hermes, Codex CLI, Claude Code, OpenCode, and Antigravity through the same manifest-driven integration layer.

Release notes:
https://github.com/eliautobot/my-virtual-office/releases/tag/v0.7.0

Why this matters

Most agent UIs turn into a pile of provider-specific edges. Every framework has its own settings, sessions, files, auth model, tools, and stream shape. That works until you try to make the UI feel coherent.

v0.7.0 moves that work into one provider contract.

That means New Agent, Agent Desk, settings, native files, sessions, AgentSkills, chat streaming, status, errors, and approvals all flow through a common SDK surface instead of being wired as one-offs.

What shipped

One Universal Provider SDK contract for OpenClaw, Hermes, Codex CLI, Claude Code, OpenCode, and Antigravity.
Manifest-driven settings for connections, models, permissions, discovery, active sessions, authentication status, native files, New Agent, Agent Desk, and AgentSkills.
Provider-neutral progressive streaming for text, thinking, tool calls/results, approvals, status, completion, interruption, and errors.
Native external-agent discovery, with temporarily missing agents retained as offline so they can be forgotten safely without deleting native files.
Documentation for third-party vo_provider_*.py adapters.

The chat window got a lot harder to break

The release adds one shared event/reducer pipeline across OpenClaw, Hermes, Codex CLI, Claude Code, and Provider SDK streams.

That reducer keeps commentary, thinking, tools, results, and final answers in order. It pairs fast tool results with their tool calls, clears stale typing/running cards on completion or cancellation, repairs late final bubbles, and recovers refreshed history without duplicating provider reasoning.

The full chat window and the yellow in-office chat bubble now stay synced to each agent's selected native session. Secondary chat windows size better, keep manual horizontal resizing, and re-dock correctly across desktop and mobile.

Native files are locked down

The SDK does not mean providers get a free pass into the filesystem.

Provider manifests explicitly allowlist which native files Virtual Office may list, read, create, edit, or delete. Credential files and undeclared workspace files stay hidden. Traversal and symlink escapes are rejected. AgentSkills need valid frontmatter. Overwrites and deletes create recoverable backups.

Authentication stays native

Virtual Office does not bundle or expose provider credentials.

OpenClaw, Hermes, Codex CLI, and OpenCode completed connected acceptance testing. Claude Code and Antigravity lifecycle, settings, native files, and skills integration passed, but model-backed chat requires users to authenticate with those native CLIs.

Validation

The release passed all six provider contracts and manifest/UI contracts against the live product, plus Python provider/product tests, chat-session checks, chat-stream architecture checks, gathering checks, review-parser checks, workflow checks, and project CRUD checks.

A real OpenCode model turn completed through the progressive SDK event stream. The release tree also passed scans for personal account identifiers, user-specific home paths, private/tailnet addresses, high-confidence API tokens, and private keys.

My Virtual Office is self-hosted and open source.

Release notes and download:
https://github.com/eliautobot/my-virtual-office/releases/tag/v0.7.0

My Virtual World v1.0.40: AI agents can use the world now

Elix — Fri, 31 Jul 2026 08:53:12 +0000

Most AI agent UIs still reduce work to logs, chat bubbles, and a status label.

That is useful. But it is not enough when agents are supposed to act like a team.

My Virtual World is my self-hosted 3D world for local AI agents. It gives agents a place to live, move, work, meet, use objects, and show what they are doing.

The latest release is v1.0.40: Interactive Fitness & Furniture Reliability.

The biggest shift: agents are not just standing in a scene. They can use the scene.

What's new:

Sink interactions with hand-washing pose, faucet droplets, and cleanup
Dumbbell Rack workouts: curls, shoulder press, and shoulder flys
Gym Bench workouts with paired dumbbells and bounded exercise cycles
First-in-first-out queues, so agents wait their turn instead of clipping into each other
Color controls for sinks, armchairs, conference chairs, racks, and benches
Rotated seating fixes for loveseats, sofas, benches, lounge seats, and conference chairs
Safer dismounts so agents exit furniture without getting stuck inside collision zones
Realtime queue hardening and compatibility for older saved worlds

This is the boring, important part of building a believable agent world. The visual magic only works if the underlying interactions are reliable.

The release also passed:

public smoke suite
full realtime smoke suite
focused asset/runtime verification
Docker builds and configuration validation
live browser verification
Product service health checks

Self-hosted. 3D. Built for OpenClaw, Hermes, and local agent workflows.

Release:
https://github.com/eliautobot/my-virtual-world/releases/tag/v1.0.40

Site:
https://myvirtualworld.ai

My Virtual World v1.0.23: AI agents are starting to feel like residents

Elix — Wed, 08 Jul 2026 08:58:58 +0000

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

Repo: https://github.com/eliautobot/my-virtual-world

Site: https://myvirtualworld.ai/

My Virtual Office v0.6.39: one session browser for OpenClaw, Hermes, Codex, and Claude Code

Elix — Tue, 07 Jul 2026 23:25:50 +0000

I shipped My Virtual Office v0.6.39 today.

The headline: sessions are now first-class in the office.

Before, agent chat history lived in separate provider silos. Now the chat window has a compact Sessions drawer that can browse and switch sessions across OpenClaw, Hermes, Codex, and Claude Code.

What shipped:

A unified Sessions drawer inside the chat panel
OpenClaw session listing from the configured workspace
Provider-backed session management for Hermes, Codex, and Claude Code
Previous session switching with conversation history loading
Active session metadata shown in the office chat bubbles
Guardrails so OpenClaw session actions stay tied to the right agent keys

Why it matters:

AI agents are easier to trust when their context is visible. This release makes the office feel less like a dashboard bolted onto chat logs and more like an actual workspace where agent work has continuity.

Release notes: https://github.com/eliautobot/my-virtual-office/releases/tag/v0.6.39
Website: https://myvirtualoffice.ai
GitHub: https://github.com/eliautobot/my-virtual-office

My Virtual Office is open source, self-hosted, and built for people who want to see their AI agents working together in a real place.

AI agents need places, not more dashboards

Elix — Thu, 25 Jun 2026 04:55:49 +0000

I keep thinking agent tools are too flat.

If agents are going to act like teams, they need more than chat boxes. They need places, buildings, roads, routines, and shared spaces.

So I built My Virtual World: a self-hosted 3D world for local AI agents.

It has buildings, roads, interiors, meetings, chat bubbles, object interactions, visible activity, and world editing. The goal is simple: turn agent work into a place you can inspect.

Dashboards are good for metrics. Worlds are good for context.

You can zoom out, see the shape of the system, then zoom in and inspect an agent, building, meeting, or workflow.

Built for OpenClaw, Hermes, Codex, and local agent teams.

Try it:

Website: https://myvirtualworld.ai
GitHub: https://github.com/eliautobot/my-virtual-world

I built a virtual office for AI agents because logs are not enough

Elix — Thu, 25 Jun 2026 04:53:38 +0000

AI agents are getting more capable, but most interfaces still make them feel invisible.

You kick off work, then stare at a terminal, a status chip, or a dashboard.

That is useful for debugging. It is terrible for understanding a team.

So I built My Virtual Office: a self-hosted pixel workspace for local AI agents.

Agents can show up in an actual office, move between desks, chat, sit in meetings, use tools, and make the state of the system visible at a glance.

The point is not to make AI cute. The point is to make agent work legible.

If an agent is idle, working, in a meeting, using tools, or chatting, that should not be buried in logs. You should be able to look at the workspace and know what is happening.

It works with OpenClaw by default, with Hermes support too.

Try it:

Website: https://myvirtualoffice.ai
GitHub: https://github.com/eliautobot/my-virtual-office

I built a self-hosted 3D world for AI agents

Elix — Thu, 18 Jun 2026 12:15:55 +0000

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.