Hermes Agent Browser: Setup Guide + Every Command

Hermes Agent browser is the built-in browser automation toolset that lets Hermes Agent open a real web page, read it, click buttons, type into forms and take screenshots for you.

It ships inside Hermes Agent, the free open-source agent from Nous Research, so you don't install a separate product to get it.

This post is the operator walkthrough, so you'll get what the toolset is, the commands that switch it on and off, every browser backend the official docs list, how logged-in sessions work, the safety settings, the costs, the limits and the fixes for the errors people hit most.

Every command and setting below comes from the official Hermes browser docs, which I re-read on 12 October 2026, plus the read-only help output of my own Hermes install on version 0.21.4.

What The Hermes Agent Browser Is In Plain English

The docs call it a full browser automation toolset with multiple backend options.

In plain English, it gives the model a browser it can drive with tool calls instead of a mouse.

The clever part is how the model sees a page.

Hermes turns each page into an accessibility tree, which is a text snapshot of the page.

Every button, link and input in that snapshot gets a short reference such as @e1 or @e5.

The model reads the snapshot, picks a reference and calls a tool to click it or type into it.

When a page only makes sense visually, the model can ask for a screenshot and have a vision model describe it.

What The Browser Toolset Can Do

The docs list these built-in tools, and I've kept to that list.

Tool	What it does
`browser_navigate`	It opens a URL, and it has to be called before any other browser tool.
`browser_snapshot`	It returns the text snapshot of the page with the element references.
`browser_click`	It clicks an element by its reference.
`browser_type`	It clears an input field and types new text into it.
`browser_scroll`	It scrolls the page up or down.
`browser_press`	It presses a key such as Enter, Tab or Escape.
`browser_back`	It goes back one page in the history.
`browser_get_images`	It lists the images on the page with their URLs and alt text.
`browser_vision`	It takes a screenshot and has a vision model analyse it.
`browser_console`	It reads console messages and JavaScript errors, and it can evaluate a JavaScript expression.
`browser_cdp`	It sends raw Chrome DevTools Protocol commands when a CDP endpoint is connected.
`browser_dialog`	It accepts or dismisses a native alert, confirm or prompt box.

The toolsets reference also lists five browser_vault_ tools that fill saved logins into sign-in forms.

There is one more tool called browser_exec, and it matters enough to get its own section.

The Default Mode Most Guides Miss

The docs say Browser Use mode is now the default browser mode.

When the browser-use command-line tool can run on your machine, Hermes gives the model a single browser_exec tool instead of the list above.

In that mode the model writes short Python that clicks, types, drags and scrapes inside the browser.

If that command-line tool can't run, Hermes falls back to the built-in tools automatically.

The docs add a safety rule here, which is that browser_exec is only offered to sessions that also have terminal access.

You can force the built-in tools yourself by typing /browser use off in the terminal chat.

I explain the Browser Use project itself in my Hermes Agent browser use guide, so this page stays on the toolset.

Browser Tools vs Web Search And Extract

People mix these up, so here is the difference.

	Web toolset	Browser toolset
Tools	`web_search` and `web_extract`	`browser_navigate`, `browser_snapshot` and the rest
What it returns	Search results and extracted page text	The live page, with clickable references
Can it click or type	No, it can't.	Yes, it can.
Speed and cost	The docs call it faster and cheaper.	It is slower, because each step is a tool call.
Best for	Looking things up	Forms, logins, buttons and dynamic pages

The docs put it in one tip, which is to prefer web search or extract for simple information retrieval.

They say to use the browser when you need to interact with a page.

🔥 Want to learn how to run a Hermes agent properly, not just install it? Inside the AI Profit Boardroom, you get step-by-step Hermes tutorials, my installable Agent OS and four coaching calls a week where you can ask setup questions live. You'll be learning with 3,400+ members. → Get access here

How To Enable The Hermes Agent Browser Step By Step

These are the real commands, and I checked each one against hermes tools --help on my own machine.

Run hermes tools list to see every toolset and whether it is enabled.
Run hermes tools enable browser if the browser line shows as disabled.
Add --platform telegram to that command when you want the change on one platform only, because the default platform is the terminal.
Run hermes tools with no subcommand to open the interactive menu, then choose Browser Automation to pick where the browser runs.
Run hermes pm install agent-browser if you installed Hermes with the skip-browser flag or the browser download failed.
Start a chat with hermes and give it a small job, such as opening a page and telling you the main headline.

When I ran the list command on 12 October 2026, the browser toolset showed as enabled for the terminal on my install.

The toolsets reference says the default terminal toolset includes the browser, so most people only need to check rather than change.

If you haven't installed Hermes yet, start with my one-click Hermes Agent setup guide and come back here.

How To Disable The Browser Toolset

Switching it off is one command.

hermes tools disable browser
hermes tools disable browser --platform telegram

Inside a chat, the same thing is /tools disable browser.

The toolsets reference makes one point worth knowing, which is that web_search is deliberately not part of the browser toolset.

That means disabling the browser doesn't take quick lookups away from your agent.

Every Browser Backend The Docs List

A backend is simply where the browser runs.

Backend	Where it runs	What the docs say about it
Local browser	Your machine	Hermes launches its own packaged Chromium through the `agent-browser` tool, with no cloud key.
Your own browser over CDP	Your machine	`/browser connect` attaches to a running Chrome, Brave, Chromium or Edge.
Real profile browsing	Your machine	Hermes drives a copy of your active browser profile, so your logins come with it.
Browser Use cloud	Cloud	It is managed Chromium with stealth, residential proxies, CAPTCHA solving and reusable profiles.
Browserbase	Cloud	It is an alternative cloud provider with anti-bot tooling.
Firecrawl	Cloud	It offers cloud browsers with built-in scraping.
Camofox	Your machine, in Docker	It is a self-hosted Firefox-based browser built for anti-detection.
Lightpanda	Your machine	It is a lightweight headless engine with no graphical renderer.
Nous subscription	Cloud	Paid Nous Portal subscribers get cloud browsing through the Tool Gateway with no separate key.

You pick the backend in hermes tools under Browser Automation.

Cloud providers need their key in your Hermes env file, and the docs are clear that a key only supplies credentials.

Once you've made a selection in the menu, adding or removing a key doesn't switch providers by itself.

I rank all of these options against each other in my best browser for Hermes list, so I won't pick a winner here.

How To Connect Your Own Chrome With /browser connect

This is the route for people who want to watch the agent work in a real window.

/browser connect
/browser connect ws://host:port
/browser status
/browser disconnect

The first command connects to a Chromium-family browser at port 9222 on your own machine, and it tries to launch one if nothing is listening.

The docs warn that this is a terminal-only slash command.

If you type it into Telegram, Discord or a web chat, it is sent to the model as plain text and nothing connects.

There is also a Chrome change you need to know about.

Since Chrome 136, the remote debugging flag is ignored when it points at your default profile folder.

The fix in the docs is to start the browser with a dedicated profile folder, like this on a Mac.

"/Applications/Google Chrome.app/Contents/MacOS/Google Chrome" \
  --remote-debugging-port=9222 \
  --user-data-dir="$HOME/.hermes/chrome-debug" \
  --no-first-run \
  --no-default-browser-check &

That dedicated profile starts signed out of everything, which leads to the next section.

Logged-In Sessions And Cookies

By default, local browsing runs in a clean throwaway profile, so the agent is logged into nothing.

The docs give you three main ways to change that.

Real profile browsing copies your active browser profile into a managed snapshot and drives that copy, so your existing logins and cookies are available.
The credential vault saves a login the first time the agent reaches a sign-in page, then fills it on later visits.
Browser Use cloud supports persistent profiles that keep cookies and saved passwords across sessions.

Real profile browsing is one setting.

# ~/.hermes/config.yaml
browser:
  use_real_profile: true

The docs say Hermes never opens your live profile directly, because it works on a separate snapshot folder.

It supports Chrome, Edge, Brave and Chromium, and it stops with a clear message if your default browser is Firefox.

The snapshot browser runs headless by default, so no window pops up while it works.

In practice you should quit your browser before starting a real-profile session, because a running Chrome usually holds the login databases with a write lock.

On Windows the browser has to be fully closed, including any tray instance.

Approvals And Safety Settings

The docs frame real profile browsing as a consent-gated convenience, not an isolation boundary.

Any page the agent visits in that mode runs with your real logins, so I'd only switch it on for a job that truly needs it.

Hermes has several guards you should know about.

The credential vault types passwords straight into the page, and the docs say the model never sees them.
A saved password is only filled when the page origin exactly matches the saved origin.
The website blocklist in the security settings blocks listed domains across web search, web extract and browser navigation.
URL checks block private and internal addresses on cloud backends, and a local sidecar browser handles localhost pages instead.
The browser.restrict_evaluate setting adds a denylist over sensitive JavaScript such as cookies, storage and network calls.

The docs are honest about the gap as well.

Once a password is typed into a site, that site has it, exactly as when you type it yourself.

On a cloud backend, the vendor's browser sees the page like any other browser would.

🔥 Want my full Hermes setup, with people to ask when you get stuck? The AI Profit Boardroom includes my Agent OS, Hermes walkthroughs and four live coaching calls a week with 3,400+ members. → Join the AI Profit Boardroom

What The Hermes Agent Browser Costs

Hermes Agent is free and open source, and the browser toolset is part of it.

Local browsing and /browser connect cost nothing beyond your model tokens.

The docs say cloud sessions consume provider credits, so Browser Use, Browserbase and Firecrawl bill you on their own plans.

The Tool Gateway route needs a paid Nous Portal subscription.

The hidden cost is tokens, because every snapshot is text your model has to read.

Limits You Should Know First

These come straight from the limitations in the docs and the notes around them.

The browser works from the text snapshot, not from pixel coordinates.
A snapshot longer than 15,000 characters is cut at a line boundary, and the full version is saved to a cache file the agent can read.
The browser can't download files.
Idle sessions are closed after 120 seconds by default.
Lightpanda has no graphical renderer, so screenshots aren't available on that engine in the default mode.

CAPTCHAs and bot detection deserve a straight answer.

The docs say Browser Use cloud includes CAPTCHA solving, and they list CAPTCHA solving under Browserbase's basic stealth.

They also describe headed mode, where the window stays open so you can solve a challenge yourself.

Nothing in the docs promises that every site will let an agent in, and my own notes agree.

One of my Hermes skills records that Product Hunt showed a "Just a moment" bot check, and the instruction I wrote was to skip that site.

Troubleshooting The Hermes Agent Browser

What you see	What it usually means	The fix from the docs
`/browser connect` gets connection refused	Chrome ignored the debug flag on your default profile.	Start Chrome with a dedicated `--user-data-dir` as shown above.
An "Allow remote debugging?" pop-up keeps appearing	You are on Chrome's approval path for your real profile.	Use a dedicated profile folder, or use real profile browsing instead.
`/browser connect` does nothing in Telegram	It is a terminal-only command.	Run it in `hermes` or `hermes chat` in the terminal.
The agent opens signed out with real profile on	Your browser was running and held the login databases.	Quit the browser fully and start a new session.
A provider error about a missing key	The selected provider has no key.	Run `hermes tools` and re-select the provider you want.
No browser tools appear at all	The local browser never installed.	Run `hermes pm install agent-browser`.
Hermes in WSL2 can't reach Windows Chrome	The debug endpoint isn't reachable from WSL.	The docs recommend the chrome-devtools-mcp server through Hermes MCP.

Real Examples From My Own Hermes Setup

I want to be exact about what my own content shows.

The clearest example is a Hermes skill I use for a daily AI news scan, which I published in my Hermes AI skills guide.

That skill tells the agent to try web search first and web extract second.

If both come back empty, it switches to browser_navigate followed by browser_snapshot with the full option on.

My note in that skill says the web tools returned nothing in March 2026 and the browser tools returned the full pages every time.

It also uses browser_scroll to reveal more of a long article.

The video at the top of this page shows a different Hermes feature, which is Computer Use.

Computer Use drives desktop apps such as Notes in the background, and the Hermes docs say browser work goes through the browser toolset instead.

I rank the browser against the other built-in tools in my best Hermes Agent tools list.

If you'd like help applying agents to your own SEO, you can book a free SEO strategy session and we'll map out a plan together. If you're already experienced and want a higher-level room, have a look at the SEO Elite Circle.

FAQ: Hermes Agent Browser

What is the Hermes Agent browser?

The Hermes Agent browser is the built-in browser automation toolset in Hermes Agent.

It lets the agent open pages, read them as text snapshots, click, type, scroll and take screenshots.

How do I enable the browser in Hermes Agent?

Run hermes tools list to check it, and run hermes tools enable browser if it shows as disabled.

Run hermes tools and choose Browser Automation to pick where the browser runs.

Is the Hermes Agent browser free?

Yes, the toolset is free and local browsing costs nothing beyond your model tokens.

Cloud browser providers use their own credits, and the Tool Gateway needs a paid Nous Portal subscription.

Can Hermes use my logged-in Chrome?

Yes, real profile browsing drives a copy of your active profile, so your logins come with it.

It is off by default, and the docs say you should only enable it when you want the agent acting as you.

Can the Hermes Agent browser solve CAPTCHAs?

The docs say Browser Use cloud includes CAPTCHA solving and Browserbase lists it under basic stealth.

Nothing guarantees every site will let an agent through, and headed mode lets you solve a challenge by hand.

What is the difference between browser tools and web search in Hermes?

Web search and extract fetch information quickly and cheaply.

Browser tools drive a live page, so you use them when the job needs clicking, typing or a logged-in session.

Also On Our Network

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

🆓 Get a FREE AI Course + Community + 1,000 AI Agents 👉

About Julian

I'm Julian Goldie, an SEO entrepreneur, author and founder of the AI Profit Boardroom, which has 3,400+ members.

I help business owners scale with AI agents, automation and SEO.

I've built a 7-figure SEO and link-building agency, Goldie Agency, with 70+ people in Bangkok.
I've grown a YouTube channel to 400,000+ subscribers.
I wrote the book "Link Building Mastery", which covers nine link building strategies.
My Udemy courses have taught 29,000+ students.

→ Get my best AI training inside the AI Profit Boardroom

Check the toolset, start on the local browser, keep real logins off until a job needs them, and you'll get real work out of the Hermes Agent browser.