Your Hermes agent office is now a real place — Hermes 3D drops your whole bot roster into a live 3D office where you watch them work and catch who’s stuck. In this guide: what shipped in v1.0, the setup tips that save an hour, and the security step most people skip.
Short answer
- Hermes 3D v1.0 (20 August): a live 3D office for your agents, running on your machine.
- Built by the Claw 3D developer — community project, MIT, 100+ stars in days.
- Standups with GitHub/Jira, PR review, QA logs, a gym for skills and a janitor for resets.
- Start in demo mode, and set the access token + allow-lists before exposing anything.
The problem a Hermes agent office solves: a team you can’t see
Bot Mode gave a lot of people a small agent team almost overnight — named bots with their own roles, models, memory and skills, messaging each other, handing work off with an @mention, running routines on their own.
And that created a new problem. A team you can’t see is not really a team. It’s ten invisible processes hiding inside terminal tabs — no idea who’s working, who’s blocked, or who handed what to whom.
Hermes 3D is the fix: a live 3D office for your agents, running on your own machine. Instead of watching them through logs and dashboards, you walk through an office where they sit at desks and work. You see who’s busy, who’s stuck, open the Kanban board, walk up to an agent and talk to it — and when one hands a task to another, you watch it happen in real time.
Where the Hermes agent office came from
The developer goes by Luke the Dev, and this isn’t his first run at the idea. He built Claw 3D on top of OpenClaw first and it went viral — then moved his own workflows to Hermes, so he rebuilt the idea properly. In his words: Claw 3D was the prototype, Hermes 3D is the real vision.
Version 1.0 went public on 20 August, open source under MIT, and passed 100 stars and around 20 forks within days. It’s an independent community project — not made by the teams behind the agent backends it connects to — and the repo says that plainly, which I respect.
🔥 Want this set up without the guesswork? Deciding which agent owns which job is the actual hard part — the office just makes the answer visible. Inside the AI Profit Boardroom you get the Agent OS as a ready-to-install file, a 30-day roadmap, daily tutorials the same day new tools ship, and four live coaching calls a week where you share your screen and get unstuck. → Get access here
What you can do inside the Hermes agent office
- Watch your agents work in real time in a shared 3D office.
- Run standups with agents connected to GitHub and Jira.
- Review pull requests from inside the office.
- Monitor QA pipelines and logs without leaving the workspace.
- Train agents in the gym to develop new skills.
- Reset sessions and clean context with the janitor system.
And v1.0 shipped real substance beyond the novelty: a cinematic renderer with PBR materials and a campus exterior, a dedicated Hermes floor with its own taskboard, direct connection to a Hermes backend with no adapter process at all — including over Tailscale — conversation huddles where chatting agents visibly group up, an office bridge so the space reacts to conversations you have in the desktop app or terminal, and a built-in Kanban board instead of a separate skill install.
What it is — and isn’t
Be clear on the boundary: Hermes 3D is the visualisation and interaction layer. It does not install the agent runtime for you. The repo puts it well — the gateway is just the control plane, and the office is the product.
It sits on top of three things: Hermes through the bundled gateway adapter, a direct HTTP custom runtime for your own orchestrator stack, or a built-in demo gateway with mock agents. That last one matters more than it sounds, because it’s where you should start.
Hermes agent office setup: tips that save an hour
- Start in demo mode. Do not wire a real backend on day one — run the demo gateway, get the mock office streaming, and decide if you like it before touching anything real.
- Check the basics: Node 20+, npm 10+. Half of all setup problems are an old Node version.
- Understand the two hops. Your browser talks to Studio, Studio talks to the gateway — so “local gateway address” means local to the machine running Studio, not the device you’re browsing on.
- Know the two gateway settings. One is a build-time variable that does nothing until you rebuild; the runtime one takes effect on a server restart. If you edit a setting and nothing changes, that’s why.
- Read connection errors literally. A protocol version error means the gateway is too old — update it or run the bundled adapter. A wrong-version error means a secure address hit a non-secure endpoint.
The security step most people skip: if you expose the app beyond your own machine, set the Studio access token — and in production, set the upstream allow-list so it can only proxy to gateway hosts you approve. There’s a second allow-list for the custom runtime route that falls back to the first if unset. This is your AI team’s front door. Lock it.
Honest limitations
- Voice transcription isn’t bundled.
- The immersive office and the builder are still separate stacks.
- The connection flow loads gateway details into browser memory — treat your browser as part of the trust boundary.
The repo lists all of that openly, which is exactly the behaviour you want from a community project connecting to your agents.
FAQ
What is the Hermes agent office?
Hermes 3D — an open-source, MIT-licensed 3D office where your Hermes agents visibly work at desks: you see who’s busy, who’s blocked, watch task handoffs live, and talk to agents by walking up to them.
Who built it?
A community developer known as Luke the Dev — the same person behind the viral Claw 3D, which he calls the prototype to Hermes 3D’s real vision. It’s unofficial and independent of Nous Research.
Does it run my agents?
No — it’s the visualisation layer on top of an existing setup. It connects to Hermes via a gateway adapter, a custom HTTP runtime, or a built-in demo mode.
How should I start?
Demo mode, always. Mock agents and streaming chat let you judge it before wiring anything real — then check Node 20+/npm 10+ before blaming the app.
Is it safe to expose?
Only with the Studio access token set and the upstream allow-lists configured so it can only proxy to hosts you approve. Don’t skip that step.
Why does a 3D office matter at all?
Because we’ve spent two years making agents more capable and almost nobody has made them understandable. Watching a handoff tells you more in three seconds than scrolling a terminal for three minutes.
The bottom line
The Hermes agent office — Hermes 3D — turns an invisible bot roster into a team you can actually watch: MIT-licensed, community-built, connecting straight to your Hermes backend, with standups, PR review and a built-in Kanban inside the space. Start in demo mode, lock the front door before exposing it, and enjoy the rare agent tool built for understanding rather than capability.
About Julian Goldie
I run Goldie Agency, a 7-figure SEO agency, and teach this daily on a 400K+ subscriber YouTube channel. 240+ client projects on Upwork at a 100% job-success score, 10+ years through every major Google update. My systems are in the AI Profit Boardroom; my link building book is free here.
Related reading
Last updated August 2026. This is the living guide to Hermes agent office — it gets updated as the tools change.