The Hermes Pixel Office is a free, open source dashboard that shows every Hermes agent on a machine as a pixel character at a desk, working in real time.
For anyone supervising agent work rather than running a single job, that changes the day quite a lot.
This is what it is, how to set it up, how to read it, and the limits worth knowing before you rely on it.
What it is
A live visual dashboard for Hermes agents, built by Teknium — one of the people behind Nous Research, the team that builds Hermes Agent.
Free and open source, with nothing to sign up for.
Every agent session becomes a character that walks into a shared office, sits at a desk, works, and walks out when the job is finished.
The supervision problem it solves
Hermes agents work in the background: running tools, spinning up helper agents, doing real work, all out of sight.
The person responsible for the output is left staring at scrolling text, unable to answer the only two questions that matter — is anything stuck, and does anything need a human.
The Pixel Office answers both in one glance.
How to set it up
Demo mode first. A built-in demo fills the office with pretend agents so anyone can learn the interface before touching a live run. For a team, this is also the training material.
Then pick a mode. Run it as its own browser page, or as a panel inside VS Code with a “+ agent” button that starts a fresh Hermes run without leaving the editor.
Connect Hermes. Every Hermes agent on the machine shares the same office automatically — command line, chat app or scheduled job, no per-agent configuration.
Remote works out of the box, which matters when the agent stack lives on a server rather than on someone’s laptop.
How to read the room
One character per agent session.
Gold collar and name tag = a helper agent spun up by a bigger agent for a side task, with the tag showing its job. That distinction is what makes a busy room legible.
The characters act out the work:
| What you see | What the agent is doing |
|---|---|
| Typing | Writing a file |
| Book open | Reading something |
| Browsing | Using the web |
| Green monitor flicker | Running a terminal command |
| Red exclamation mark | Waiting on a human |
| Standing up, walking out | Task finished |
Approval flags, counter and chime
When an agent hits a command that could be risky, it needs a yes or no before continuing.
That character throws up a bright red exclamation mark, and a counter at the top shows how many are waiting — for example “2 waiting”.
An optional chime, off by default, dings when an agent needs approval or when a helper finishes.
For a supervisor that counter is the whole product: it converts “somebody should check on the agents” into a number that is either zero or not.
One office for the whole machine
Every Hermes agent running on the machine appears in the same room.
Several jobs at once means several characters, one window, and no per-job dashboards to keep open.
What it deliberately does not do
It does not slow the agents down. It is a window and nothing more — it draws what the agents are already doing, never blocks them, never changes their work and never adds weight to how they run.
It does not control anything. Keep the agents’ own safety rules on. The red flag tells a human when to look; the human still says yes or no.
It is not an audit trail. It is a live read on state. Keep proper logging for anything you need to review later.
Old way vs new way
| Old way | New way |
|---|---|
| Someone reads logs to see whether a batch is progressing. | Anyone glances at a room and knows. |
| Blocked agents sit unnoticed until someone checks. | Red flag, waiting counter, optional chime. |
| Sub-agent work invisible inside the parent output. | Gold collars and name tags on every helper. |
| Monitoring assumed to cost performance. | Passive window, zero impact on the run. |
🔥 Want help wiring agents into your own workflow?
Book a free AI SEO strategy session and we will look at where your delivery hours actually go: go.juliangoldie.com/strategy-session. More on how we work: goldie.agency.
Three jobs it is genuinely useful for
Batch content production. A group of Hermes agents drafting hooks, scripts and captions, each helper gold-collared and several working at once, with the red flag marking the exact moment a decision is needed.
Onboarding and sequence flows. Agents laying out welcome messages, call prep and resource walkthroughs — every character finishing and walking out confirms the flow completed and nothing stalled halfway.
Lead systems. Agents planning content, hooks and follow-ups, with one glance showing who is working, who finished and who is blocked on a human.
Practical tips
One. Use demo mode as onboarding for anyone who will be watching the room.
Two. Enable the chime for approvals only, so the sound always means action is needed.
Three. Treat “character stopped moving” as a signal worth investigating, not a curiosity.
Four. Keep logging in place — the office is speed, not a record.
Frequently asked questions
What is the Hermes Pixel Office?
A free, open source dashboard showing every Hermes agent as a pixel character in one shared office.
How do you set it up?
Demo mode first, then a browser page or VS Code panel, then connect Hermes.
Can it approve actions?
No — it is a window, not a controller.
Does it work remotely?
Yes, out of the box.
Does it slow agents down?
No, it is entirely passive.
About Julian
I’m Julian Goldie, founder of Goldie Agency, a 7-figure SEO and link building agency with a 70+ team.
400K+ YouTube subscribers, 163K X followers, 29K+ Udemy students, and author of Link Building Mastery.
We run agent batches on live delivery work, so the supervision angle above comes from needing to answer “is anything stuck” quickly.
Also on our network
More on the Pixel Office: juliangoldie.com, juliangoldie.co.uk and goldstarlinks.com.
📺 Video notes + links to the tools 👉
🎥 Learn how I make these videos 👉
🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉
The Hermes Pixel Office does not do the work — it lets whoever is responsible finally see it.