Herald OS is a free, open-source, agent-native operating system that puts Hermes Agent at the centre of your computer, so you type or talk to an AI agent instead of clicking through menus. It is built by an independent developer, it is in early alpha, and I set it up in about two minutes. Here is the full tour of what it does, where it breaks, and who should use it.
📺 Watch: Herald OS DESTROYS Hermes Agent Desktop?
🔥 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
Quick context. When Hermes Agent first came out around March or April, pretty much everyone ran it in the terminal. It worked, but it was messy. A proper UI makes it far easier to navigate, control and command your agents. That is the gap Herald OS fills.
What is Herald OS?
Per the README, Herald OS is "an agent-native operating system, with Hermes Agent as the interface." You talk to Hermes, and Hermes opens apps, organises files, runs scheduled routines and remembers your preferences.
The key facts, all from the README at iamlukethedev/Herald-OS:
- Who makes it: Luke The Dev (iamlukethedev), who posts a lot about Hermes on X. It is an independent project.
- Not Nous Research: it is not an official Nous Research product and is not affiliated with or endorsed by Nous.
- Licence: MIT, building on Hermes Agent, which is also MIT.
- Status: alpha (0.1). The first public release, v0.1.0-alpha.1, landed on 6 October 2026.
- Traction: about 124 stars and 17 forks at the time of writing, plus a Discord community.
The README is honest about the stage it is at: "Expect rough edges, and keep backups of anything you let an agent touch." Take that seriously.
If you want to go deeper on Hermes itself, the AI Profit Boardroom has full Hermes Agent courses, daily Hermes updates and a course on building your own agentic operating system. Join the AI Profit Boardroom for $69/mo.
How I set up Herald OS
I literally went into Claude and said, can you set this up? About two minutes later it was running. That is the easiest way to do it.
If you would rather do it by hand, the README order on a Mac is simple. Install Hermes Agent first using the official install script from hermes-agent.nousresearch.com, then run hermes setup. After that, download the HeraldOS mac-arm64 .dmg from the repo's Releases page and drag it into Applications. The build is not notarised yet, so the README gives a one-time command to clear the macOS quarantine flag.
The full step-by-step, including the Linux options, is in my Hermes Agent OS GitHub install guide.
What you need, per the README:
- Hermes Agent installed separately (a source build needs Hermes Agent 0.21.3 or newer).
- A model provider: a Nous Portal account or an API key from OpenRouter, OpenAI, Anthropic, a local model or others.
- Mac: macOS 13 Ventura or later on Apple silicon only. There is no Intel build.
- Linux: an x86_64 installer ISO, an Arch package, an Omarchy app, or a tarball for other distros.
The Herald OS tour: every page I clicked
Dashboard
The first thing you see is "Good morning, Julian. Your day already in motion." There is voice input right there. It shows what is scheduled today, pulled from my scheduled tasks, what needs attention, a pick-up-where-you-left-off toggle and your missions. It even used recent screenshots from my desktop on the dashboard.
The Hermes chat page has a preview panel and sources, so you can add files Hermes can read. You can start a new conversation any time.
Spaces
You get spaces for ideas, work and personal. I really like this. AI Profit Boardroom members ask me all the time how to set up an agent OS for separate companies or clients. Here you just add a new space per client and set a default space. Done.
Missions
You give Hermes a mission, something like "build out a new website for me", and it plans it step by step. This is the bit that feels most like an operating system and least like a chatbot.
Memory
The Memory page shows everything Hermes remembers about you. Fair warning: mine showed my Telegram details on screen. If you screen-share or record videos, be careful what you have open here.
Files
The Files page has recent, favourites, projects, downloads, shared and deleted. Clean and simple.
Automations and sync
You can switch each automation on or off. The part that impressed me: my automations synced automatically across Herald OS, Hermes Desktop and my own Agentic Operating System, which also manages agents beyond Hermes. My examples were a daily 6am to-do recap, daily Japanese learning, and morning inbox triage summaries.
Per the README, automations can also trigger on login, after a break, on low battery or after a crash.
Connections
This is where I hit a wall. The Connections page did not work or sync for me. My plan is to ask Claude to fix it. It is open source, so who am I to be picky?
Settings
Settings is huge. Here is what I went through:
- Preferred model.
- Where agents complete tasks, including Docker, which people ask me about a lot.
- Number of background agents.
- Remember preferences.
- Auto-approve on or off, for example "ask before sending messages".
- Activity history. Useful, because stuff gets set up in the background and you forget about it.
- Launch full screen (I do not), default space and default projects folder.
- Do not disturb, which silences the bell sound, plus stay awake, screen saver and dictation.
- Colours and themes. I like the default.
- Software it can use, such as Claude Code, Codex and LM Studio.
- Plugins: paste a GitHub link and click add.
- A Hermes agent section and a usage page with tokens used, plan and balance.
- Privacy, notifications, network, storage and accessibility.
Per the README you also get twelve themes (two light), a theme from an image, themes installed from git and custom fonts. The usage page tracks weekly and monthly usage and warns you at 90%.
Voice
The voice settings let you pick a local or live engine, test it, and see a list of voice commands. Per the README, you press the voice key or say "hey Hermes", and you can drive the UI by voice, for example "open missions".
Extras from the README
- A Terminal, System monitor and floating chat windows.
- A command bar to open pages, add memories, run automations and start missions.
- Studio, which builds a project from a description with a live preview.
- Crash help, sandboxed widgets and screen tools: ask about an area of the screen, a colour picker, QR, screenshots and recordings.
- On Linux: a control menu, one-click installs for AI tools, Wi-Fi, Bluetooth, sound and display panels, clipboard history, a firewall, plus herald-os update and herald-os rollback.
Want help deciding how an agent OS fits your business? Book a free AI strategy session.
Is Herald OS safe to use?
It is reasonably careful by design, but it is alpha software, so keep backups. Per the README, every system tool has a permission tier. Read and act tools run directly and get logged. Mutate tools ask first. Destructive tools ask every time. Your ~/.ssh folder, keychains and Hermes's own credentials are always refused, and there is an audit log saved on disk.
One privacy detail worth knowing: pick up where you left off is off by default. When you switch it on, it suggests up to three threads of work using names and dates, not file contents, and those are sent to your model provider.
Rough edges I hit
- The top of the window cuts off slightly.
- The Connections page did not work or sync for me.
- No Intel Mac build, and the Mac app is not notarised yet.
- It is version 0.1 alpha. Things will change and things will break.
None of these stopped me using it. They are what you expect from a release that is days old.
Where Herald OS fits next to Hermes Desktop
I think Herald OS looks much nicer than Hermes Desktop. Hermes Desktop is still good, and it has plugins. If you want the side-by-side, read my Hermes Desktop app review. And if you want to understand the bigger idea of running Hermes as a full system, my Hermes agentic operating system breakdown covers it. For how I set up my own system, see my Agent OS guide.
Herald OS lets you choose your model with hermes model or in Settings, under Hermes and agents. If you are not sure which model to run, check Goldie Bench.
My honest verdict and who it is for
Super nice, super clean, easy to use, and it synced up pretty quickly. If you want a nice UI for Hermes Agent and you do not love Hermes Desktop, Herald OS is a great option. It is free, and it is on GitHub.
It suits you if:
- You already run Hermes Agent and want a proper visual home for it.
- You run several companies or clients and want separate spaces.
- You are on an Apple silicon Mac or Linux and comfortable with alpha software.
Skip it for now if you are on an Intel Mac, or if you need everything to work perfectly on day one.
If you want the Hermes courses, the daily updates and the agent OS build course in one place, get inside the AI Profit Boardroom. If you want a plan for your own business, grab a free AI strategy session.
FAQ
What is Herald OS?
Herald OS is a free, MIT-licensed, agent-native operating system that uses Hermes Agent as the interface. You type or speak to Hermes, and it opens apps, organises files, runs routines and remembers your preferences.
Is Herald OS made by Nous Research?
No. Herald OS is an independent project by Luke The Dev. Per the README, it is not an official Nous Research product and is not affiliated with or endorsed by Nous. It builds on Hermes Agent, which Nous Research makes.
Is Herald OS safe?
It has sensible guardrails: permission tiers, approval prompts before changes, refused locations like ~/.ssh and keychains, and an audit log on disk. It is still alpha, so keep backups of anything you let an agent touch.
Does Herald OS cost anything?
Herald OS itself is free and open source on GitHub. You still need a model provider, either a Nous Portal account or an API key, and that usage is what you pay for.
