TopGit
GitHub Repo Review

My Virtual Office: A 2D Workspace for AI Agents

eliautobot/my-virtual-office
MTopGit review image for eliautobot/my-virtual-office
Review by Topgit.dev for eliautobot/my-virtual-office, with GitHub repository stats and README context.
Quick verdict

My Virtual Office is a self-hosted 2D workspace that turns your AI agents into characters walking around an office you can watch and click into. It connects to OpenClaw, Hermes, Codex, and Claude Code, syncing agent activity onto the canvas live. Reach for it if you run several coding agents and want one status view instead of scattered logs; skip it if you run one agent or don't want a browser tab devoted to watching characters walk to desks.

Stars
★ 324
Forks
⑂ 74
Language
Python
License
AGPL-3.0
Topic
AI Tools
Updated
Aug 2026
Homepage
GitHub

What is My Virtual Office?

My Virtual Office is a self-hosted 2D AI workspace built in Python that renders your AI agents as characters in a browsable office scene instead of raw log output. It connects to agent harnesses like OpenClaw, Hermes, Codex CLI, and Claude Code, and shows each agent walking to a desk, sitting in a meeting, or idling in the kitchen based on what it's actually doing.

Key Features for AI Agent Visualization

  • Live Office Canvas: renders agents in real time with A* pathfinding, collision avoidance, and wall occlusion (agents behind walls get shadowed), at 100 FPS with a configurable canvas size.
  • The comprehensive office editor enables furniture placement via drag-and-drop, incorporating snap-to-grid functionality for 25+ distinct furnishings, such as desks, a meeting table, couches, vending machines, and a ping pong table; furthermore, a Layout Marketplace facilitates saving and sharing of layouts, formatted as portable .mvo-layout.json files.
  • Meeting System: agents walk to each other's desks for 1-on-1s, or gather at a 10-seat meeting table (5 per side) for group meetings, with status visible on the dashboard.
  • Branch System: agents organize into color-coded departments (Engineering, Sales, Support, etc.) with per-branch wall accent colors and emoji.
  • Dashboard Panel: live CPU/RAM monitoring, per-agent API usage/cost tracking, an agent directory with live status (working, idle, meeting, break), and an activity log.
  • Chat with Agents: markdown-rendered chat with syntax highlighting, inline images with a lightbox view, file attachments, and a tool-activity feed showing exec/read/write/search calls as they happen.
  • Dynamic environment touches: a weather-driven window view tied to your location, a day/night ambient lighting cycle, and a TV with 5 channels agents walk over to watch.
  • Office Pet: a Cat, Pug, or Lobster (default named Clawy) with its own pathfinding, that agents can pet or chase around the room.
How this repository's GitHub stars have grown over time. Source: star-history.com.View the star history

How to Use My Virtual Office

  • Running several coding agents (say, one Claude Code, one Codex, one OpenClaw agent) side by side and seeing which one is actually working versus idle without tailing three terminal windows.
  • Grouping agents into branches like Engineering or Support so a team of agents reads like an org chart instead of a flat list.
  • Reviewing an agent's tool-activity feed (exec, read, write, search) mid-task from the chat window instead of scrolling raw logs.
  • Tracking API usage and cost per agent from the Dashboard Panel across a session.
  • Scheduling recurring agent tasks visually with the Cron Manager (paid feature) instead of hand-writing cron syntax.

Installing My Virtual Office with Docker

Clone the repo, then run docker compose up -d from the project directory — that's the README's whole documented path. Once the container is up, open http://localhost:8090/setup to run the setup wizard, connect OpenClaw and/or Hermes, and either enter a license key or continue in demo mode. Prebuilt images are published at ghcr.io/eliautobot/my-virtual-office, tagged latest, by release version (for example 0.7.3), or by commit SHA prefix for exact rollback pins, with multi-arch builds for linux/amd64 and linux/arm64 — the README notes this covers both standard x86_64 machines and a Raspberry Pi 5. Updating is docker compose pull followed by docker compose up -d; your license key, office layout, and settings persist in the vo-data volume across updates.

Strengths

  • Turns agent activity into something glanceable — a live office view with a status-tagged agent directory instead of tailing multiple terminal windows.
  • Covers four real coding-agent harnesses out of the box (OpenClaw, Hermes, Codex CLI, Claude Code), each wired to its native protocol rather than a generic wrapper — Claude Code chat, for instance, uses the CLI's own stream-json protocol.
  • Docker Compose is the whole documented install path, and multi-arch images (linux/amd64, linux/arm64) mean the same tag runs on a regular server or a Raspberry Pi 5.
  • The free demo is a real, no-signup way to see the office canvas, branch management, and chat with up to 3 agents before deciding whether to pay.
  • The README is unusually specific about per-harness wiring — environment variables, mount paths, and session-hydration behavior are documented per integration rather than glossed over.

Free Demo vs. Full License Features

  • The free demo caps you at 3 agents, shows a watermark and demo banner, and locks out the office editor, agent customization, office pet, Agent Browser, SMS/Twilio panel, Cron Manager, and Whisper voice input — most of the actual office experience sits behind a paid license key.
  • License keys are validated once online through Lemon Squeezy before they work offline, so first activation needs internet access even though the app is otherwise self-hosted.
  • The core app is AGPL-3.0-or-later, but the premium feature set and hosted distribution are commercial — a network-hosted modified version still triggers AGPL's source-disclosure requirement, so this isn't a simple open-source-with-no-strings project.
  • The README documents no built-in authentication for the web UI itself; it instead tells you to keep the app on a private LAN or behind Tailscale, so securing remote access is entirely on you.
  • This is a young, narrowly-adopted project — the README gives no indication of how many people run it in production or how mature the premium feature set is.
  • Agent harness support is specific: OpenClaw, Hermes, Codex CLI, and Claude Code are documented as native integrations, plus OpenCode and Antigravity via the Provider SDK — anything outside that list isn't covered.

Alternatives to My Virtual Office

Claude-Code-Agent-Monitor — a dashboard built specifically for watching Claude Code agent activity, log-and-status first rather than a 2D office you walk around in.agent-orchestrator — handles routing and coordination across multiple agents without the visual office metaphor My Virtual Office is built around.Claude Squad — a terminal-based tool for running and switching between multiple coding-agent sessions (Claude Code, Codex, and others) in tmux panes instead of a graphical office.Vibe Kanban — organizes agent work as tasks on a kanban board rather than characters walking around a virtual office.

My Virtual Office FAQs

What agent harnesses does My Virtual Office support?

My Virtual Office natively supports OpenClaw, Hermes Agents, Codex CLI, and Claude Code, plus OpenCode and Antigravity through its built-in Provider SDK, according to the README.

Is My Virtual Office free to use?

My Virtual Office runs free in Demo mode, capped at 3 agents with a watermark and demo banner; a paid license key enables unlimited agents and the rest of the feature set.

What is the license for My Virtual Office?

My Virtual Office licenses its code as AGPL-3.0-or-later — the GNU Affero General Public License, version 3 or any later version; a separate paid license key activates the hosted/product feature set on top of that code license.

How do I secure My Virtual Office for remote access?

My Virtual Office's README recommends keeping it on your local machine, LAN, or a private Tailscale tailnet rather than port-forwarding 8090, 8091, or your agent-harness gateway ports to the open internet.

Can I customize the office layout and agents?

My Virtual Office includes a drag-and-drop office editor with 25+ furniture items, wall and floor color options, and a full agent appearance editor for skin tone, hair, costumes, and held items — though the editor and appearance tools require a paid license.

Does My Virtual Office support voice input for chat?

My Virtual Office supports voice input for chat through Whisper speech-to-text, listed in the README as a premium feature that requires a paid license key.

The problem it solves

Running multiple coding agents (a Claude Code session here, a Codex CLI agent there, maybe an OpenClaw agent too) usually means several terminal windows or scrollback logs with no shared view of who's doing what right now. My Virtual Office's README frames its own reason for existing around exactly that gap: agent work is normally invisible, so it gives each agent a visible presence — walking to a desk, sitting in a meeting — in a shared 2D office instead.

Who should try it — and who should skip

Try My Virtual Office if you're already running several agents through OpenClaw, Hermes, Codex CLI, or Claude Code and want one browser tab that shows who's working, idle, or in a meeting instead of switching between terminals. Skip it if you run a single agent, need a browser tab that isn't hosting a background office scene, or aren't ready to pay for a license once you outgrow the 3-agent demo cap.

Related repositories

Source & attribution

Facts sourced from the eliautobot/my-virtual-office GitHub repository and its README.

GitHub data · last synced Aug 12, 2026Reviewed by Henry
Back to TopGit

Still deciding about my-virtual-office?

One click hands the question to an AI along with this page — see what it says about my-virtual-office.

GitHub