The Hermes agent robot is the Autonomous Lamp, a desk-lamp robot that gives Hermes a body. It's a fully assembled, open-source companion robot from Autonomous, and it hands your agent eyes, ears, a voice and a body it can actually move.
📺 Watch: Hermes Agent + Autonomous Lamp is INSANE
🔥 Get the Agent OS as a free bonus: AI Profit Boardroom members get the full Agent OS zip, prompt libraries, daily tutorials and weekly live coaching calls. → Get inside
Autonomous announced it with one line: "Hermes just escaped. We just gave Hermes a body, one eye, and two ears." That's the whole pitch. Your agent stops living only in a terminal or a chat window and starts sitting on your desk.
I run Hermes every day inside my Agent OS. I made a video on this lamp on 23 October 2026. Here's what it does, straight from Autonomous's product page and the lampOS README.
If you want the full Hermes setup I use daily, it's inside the AI Profit Boardroom for $69/mo.
What Is the Hermes Agent Robot?
Autonomous calls the lamp "your personal robot". It's shaped like a desk lamp, but it's a robot first. It has a camera for an eye, a microphone array for ears, speakers for a voice, and servos so it can move and look around.
Hermes isn't the only brain it supports. The lamp works with six agents:
- Hermes
- Claude Code
- Codex
- OpenClaw
- OpenCode
- PicoClaw
So think of it as a body your agent can borrow. You've seen what Hermes can do on a screen with Hermes Cua. This is the same idea pointed at the physical world.
Autonomous Lamp Specs
Here's what's inside, straight from the product page.
| Part | Spec |
|---|---|
| Movement | 5 position-feedback servos |
| Camera | Wide-angle, with face tracking and motion sensing |
| Audio | Built-in microphone array and stereo speakers |
| Processor | 8-core ARM64 CPU |
| Memory | 6 GB RAM |
| Air sensors | CO₂, airborne particles (PM1.0 to PM10), temperature, humidity |
| Power | 12V 5A DC |
| Size | 7.87 × 7.87 × 18.43 inches |
| Weight | 6.6 lbs |
| Warranty | 1 year |
The product page lists it at $999 for one, with lower per-unit prices if you buy three or more.
📺 Watch: Hermes3D Is INSANE: AI Agents Now Have a 3D Office
Choosing the Brain
You pick what powers it. Two options:
- Cloud model: from the provider you choose.
- Local model: running on your own hardware.
Then you pick the agent engine on top. All six agents sit behind one gateway. You can switch between them live from the web UI, and the persona, memory and connectors move with it. So you could run Hermes in the morning and swap to Claude Code in the afternoon without rebuilding anything.
You also pick its voice from a supported speech service and shape its personality. More on that when we get to SOUL.md.
Not sure which model to run as the brain? My Goldie Bench results are a good place to start.
What You Can Do With It
The product page lists a lot. Here's the short version:
- Talk to it. Normal conversation, like you would with any agent.
- Show it things. Hold up a drawing or an object and ask about it.
- Take pictures. It can snap a photo through its camera.
- Get nudges. Posture, hydration and break routines. Camera-based skills use a supported sensing service.
- Use the ring light as a cue. The light itself becomes a signal.
- Ask about the room air. CO₂, particles, temperature, humidity.
- Connect your tools. A calendar, a project tool, or an agent on another machine.
- Run your home. Connect Home Assistant for lights and home routines.
There's also a computer-use skill that drives Mac apps through Autonomous Buddy, which bundles Cua Driver. If that sounds familiar, it's the same territory as the Hermes computer use skill. And harness-use sends coding and research tasks to agents in Autonomous's Harness on a paired computer.
Want help working out where an agent like this fits in your business? Book a free AI strategy session.
📺 Watch: Hermes AI Agents Just Went Portable
How lampOS Works
The software is lampOS. It's on GitHub at autonomous-ai/lampOS under the Apache-2.0 licence, and it describes itself as "The open-source operating system for physical AI". It had about 490 stars as of 24 October 2026.
Autonomous calls it the Android for physical AI agents. It sits between your agent and the hardware. Your agent talks to lampOS, and lampOS talks to the motors, camera, speakers and sensors.
Three files describe the robot
- SOUL.md is who it is and how it talks. Edit it and it's someone else on the next turn.
- SAFETY.md is how fast, how bright and how late. Brightness, quiet hours, move speed.
- SKILL.md is one thing it can do. A skill is one folder with one SKILL.md.
That's it. Plain text files. If you've worked with agent skills before, you already get this. My list of the best Hermes agent skills uses the same one-folder-one-skill thinking.
Skills without coding
You can install skills from the Skill Store in one tap. Or you type what you want in the app and it writes the skill for you. The demo shows a photo booth skill.
The safety gate
This is the part I like most. The safety gate is a pure function of SAFETY.md in every request path. There's no model in the loop. So the limits are the same whichever agent or model asked. Hermes can't talk its way past quiet hours, and neither can Codex.
111 endpoints
The hardware is exposed over HTTP through 111 endpoints with live Swagger docs. The video rounds it to "over 100". Either way, every part of the lamp is something your agent can call.
Home Assistant and other robots
lampOS connects to Home Assistant for lights and home routines. It also runs on other robots, such as a Reachy Mini port. Autonomous calls this "bring your own robot".
Business Uses for a Hermes Agent Robot
The product page has a "Make money" section. The idea is simple: build robot experiences for shops, teams, schools, creators and events.
Here's how I'd think about it, using only what the lamp can already do:
- Events: the photo booth skill from the demo is an obvious starting point.
- Shops: a talking, moving lamp that answers questions is a lot more memorable than a sign.
- Teams: connect a calendar or project tool and let it give updates out loud.
- Schools: show it a drawing or object and let it explain.
- Creators: a physical agent on your desk is content by itself.
Since skills are just folders with a SKILL.md, you can build a skill once and set it up for client after client. If you like the idea of Hermes making physical-world stuff, check out Hermes Meshy too.
The bigger win is still the agent behind it. I teach the full system in the AI Profit Boardroom: the Agent OS, a 30-day AI agent roadmap, daily tutorials, 4 coaching calls a week, a prompt library and a member map. 3,400+ members, and we're rated 4.9 on Skool.
Getting Started in 5 Steps
From the video, setup goes like this:
- Set it up in the Autonomous app on your phone and connect it to Wi-Fi.
- Pick your agent. Hermes, Claude Code, Codex and so on.
- Choose its voice and personality.
- Add a skill from the Skill Store, or describe a new one and let the app write it.
- Set your safety limits in SAFETY.md: brightness, quiet hours, move speed.
That's the whole flow. No soldering, no assembly. It arrives fully assembled.
If you're new to running Hermes at all, start with my Agent OS guide so the brain is set up before the body shows up.
Who It's For
This lamp makes sense if you:
- Already run Hermes or another agent and want it off the screen.
- Want an open-source robot you can actually change, not a closed gadget.
- Build experiences for shops, events, schools or creators.
- Run Home Assistant and want a voice and camera on top of it.
- Like the idea of writing a skill in plain text and seeing it move.
If you just want a chatbot, you don't need a robot. A Hermes agent bot on your phone or desktop will do the job.
Not sure which side you're on? Grab a free AI strategy session and we'll work it out.
FAQ
What is the Hermes agent robot?
It's the Autonomous Lamp, an open-source desk-lamp robot that gives Hermes and five other agents eyes, ears, a voice and a body it can control.
How much does the Autonomous Lamp cost?
The product page lists $999 for one, with lower per-unit prices for three or more.
Can I run a local model on it?
Yes. You can pick a cloud model from the provider you choose, or a local model on your own hardware.
Can the agent ignore the safety limits?
No. The safety gate is a pure function of SAFETY.md in every request path, with no model in the loop. The same limits apply whichever agent or model asks.
Want a plan for putting agents like this to work in your business? Book a free AI strategy session and get the Hermes agent robot working for you.











