fbpx

Hermes Desktop Browser: The Agent Browses For You

WANT TO BOOST YOUR SEO TRAFFIC, RANK #1 & Get More CUSTOMERS?

Get free, instant access to our SEO video course, 120 SEO Tips, ChatGPT SEO Course, 999+ make money online ideas and get a 30 minute SEO consultation!

Just Enter Your Email Address Below To Get FREE, Instant Access!

Your AI agent can now open the web itself — not a link, not a screenshot you paste in, the actual live page next to your chat. Here’s what shipped in the Hermes desktop update, what I got it to do, and three tips that save you a lot of pain on day one.

Short answer

  • A right-hand preview rail renders live pages, files and tool outputs beside your chat.
  • Shipped in v0.20 — the Herald release, 3 August — with 650+ contributors in one window.
  • Already have Hermes in the terminal? Just run hermes desktop, no separate download.
  • Leave approvals ON — it runs real commands, and browsing means reading pages you don’t control.

What actually changed

The Hermes desktop app has a right-hand preview rail sitting next to your chat. It renders web pages, files and tool outputs side by side while you keep talking to the agent.

Pair that with the browser tools Hermes already has, and the agent can open a page, move around inside it, and tell you what’s on it. You stay in one window the whole time. No tab hopping, no pasting a URL into a chat box and hoping the model reads it properly.

People are calling it the app you never have to leave. Some are reporting rough edges too, which is fair for a release this size.

That’s the real shift: the agent stops telling you to go and look at something and goes and looks at it itself.

This shipped in v0.20, the Herald release

The desktop app first appeared in June as a preview. The version everyone’s talking about is v0.20, which Nous Research call the Herald release. It dropped on 3 August.

Metric One release window
Commits ~3,650
Merged pull requests ~1,400
Issues closed ~1,200
Contributors 650+

Full changelog in the v0.20 breakdown and the release story in Hermes Herald.

🔥 Want this set up without the guesswork? The install goes fine — then the real questions hit. Which model do I point it at? Why did my session slow down? 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 I got it to do

I asked it to look up a site and it pulled the page straight into the preview rail. Then I told it to navigate inside — go to this section, then that one. It did both, and I never touched a browser tab.

In the same session I asked it to create a simple HTML landing page for a free SEO audit offer, with a headline, a benefit section, a form, testimonials, an FAQ and call-to-action buttons. It built the page.

That’s the pattern you’ll use most: describe the thing in plain English, let it build, then look at the result in the preview rail.

The rest of what shipped

Real conversational voice

Not the old talk-wait-reply style. It speaks as it generates, so it starts talking before it’s finished thinking. You can interrupt mid-sentence and it stops and listens, and it’s smart enough not to talk over you. Works across the command line, the desktop app and the messaging platforms it connects to.

Wake words

Set your own wake phrase and the detection runs on your own machine — nothing gets sent anywhere while it sits waiting. You can point different wake words at different profiles, one for work and one for personal, and just say stop to end the voice chat.

Grounded citations

A new research skill produces work with verifiable sources. Quotes get matched against the actual text on the page so it isn’t inventing them, and citations link to the exact evidence. There’s a fact-checking mode you can point at a document or a claim.

A proper platform

Generated HTML and small apps now run in a sandboxed live preview next to your chat. There’s a plugin system with a Kanban board as the first one, a global hotkey that captures a thought into the agent from anywhere on your computer, and you can run multiple windows at once.

Agent to agent

Hermes now speaks the A2A protocol v1.0. In plain English, your agent can discover other agents, talk to them, and be driven by them. If you’ve wanted specialised agents working together instead of one doing everything, that’s the door opening.

Installing it

  1. Head to the Nous Research Hermes site — there are downloads for Mac, Windows and Linux, plus a one-line terminal install.
  2. If you already have Hermes in your terminal, don’t download anything separately. Just run hermes desktop and it launches using your existing config, keys, sessions and skills.
  3. On Windows the installer isn’t code signed, so SmartScreen throws a warning the first time. Click More info, then Run anyway. That’s expected and Nous say so in their docs.

It’s the same agent everywhere — CLI, desktop and the gateway that connects it to Telegram, Discord, Slack, WhatsApp and email. Same settings, same memory, same sessions. Start something in the desktop app and pick it up in the terminal later; nothing resets.

Three tips that save you pain

  1. Use the preview rail deliberately. Don’t just let the agent describe what it did — ask it to show you. Pull the page up, look at it, give feedback on what you’re actually seeing. That one habit cuts your revision cycles right down.
  2. Watch the context meter. There’s a live meter in the status bar showing how full your session’s context is. Click it for a breakdown of what’s eating the space. When it fills up, start a fresh chat rather than pushing on — long messy sessions are where quality quietly drops.
  3. Don’t turn off the approvals. There’s a mode that skips confirmation prompts on risky commands and it feels fast, but this agent runs real commands on your real machine. Leave approvals on while you’re learning and turn them off later on purpose, when you know exactly what you’re allowing.

Bonus: type /journey in the chat and you get an interactive map of everything Hermes has learned — every skill and memory on a timeline. Scrub through it, edit or delete anything wrong. Check it weekly. Clean memory means a better agent.

One security caveat worth stating. An agent that browses live pages is reading content you don’t control, and instructions hidden in a web page can attempt to influence what the agent does — that’s prompt injection. It’s the reason tip three matters: keep approvals on, be careful what you point it at, and don’t give it access to anything sensitive while you’re still learning its behaviour.

Want it running inside a proper system? Everything I test runs inside the Agent OS, which is in the AI Profit Boardroom — walkthroughs for tools exactly like this one, weekly coaching calls where you bring your own screen and get unstuck on your config, prompt libraries you can copy straight in, and 30-day roadmaps so you know what to build in week one. Start free with the free AI course and community.

FAQ

What is the Hermes desktop browser update?

A right-hand preview rail in the Hermes desktop app that renders live web pages, files and tool outputs beside your chat, so the agent can open and navigate pages itself instead of you pasting URLs.

Which version added it?

v0.20, the Herald release, which dropped on 3 August 2026 with around 3,650 commits and 650+ contributors in one window.

Do I need to reinstall if I already have Hermes?

No. Run hermes desktop and it launches using your existing config, keys, sessions and skills.

Why does Windows warn me when installing?

The Windows installer isn’t code signed, so SmartScreen flags it. Click More info then Run anyway — Nous Research document this.

Can it really browse a site on its own?

Yes. In my test it opened a site in the preview rail and navigated to specific sections on instruction, without me touching a browser tab.

Is it safe to let an agent browse?

Keep approvals on. A browsing agent reads pages you don’t control, so hidden instructions in page content are a real risk. Don’t point it at anything sensitive while you’re learning it.

What is /journey?

A command that shows an interactive timeline of everything Hermes has learned — every skill and memory — which you can scrub through and edit or delete. Worth checking weekly.

Does the voice feature work everywhere?

Yes — the conversational voice works across the command line, the desktop app and the messaging platforms Hermes connects to.

The bottom line

The Hermes desktop browser update is the moment the agent stopped asking you to go and look at things. A preview rail beside your chat, live pages it navigates itself, voice you can interrupt, and citations it can’t fake — just keep approvals on, watch the context meter, and check what it’s learned every week.

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.

Picture of Julian Goldie

Julian Goldie

Hey, I'm Julian Goldie! I'm an SEO link builder and founder of Goldie Agency. My mission is to help website owners like you grow your business with SEO!

Leave a Comment

WANT TO BOOST YOUR SEO TRAFFIC, RANK #1 & GET MORE CUSTOMERS?

Get free, instant access to our SEO video course, 120 SEO Tips, ChatGPT SEO Course, 999+ make money online ideas and get a 30 minute SEO consultation!

Just Enter Your Email Address Below To Get FREE, Instant Access!