agentos.guide › blog

Hermes Agent Browser Use: The Quick Start, The Three Real Routes, And Where It Fits In Your Agent OS

By Julian Goldie · 12 October 2026 · agentos.guide

Hermes Agent Browser Use is a real, built-in integration: Browser Use is the default browser driver inside Hermes Agent, and Browser Use Cloud is one of the cloud browsers Hermes can use.

You do not need a plugin, and you do not need an MCP server.

Below is the quick start, the three real routes, how I'd slot it into an Agent OS, three browser tasks to copy, and the safety rules I'd set first.

I checked everything against the official Hermes docs, Browser Use's docs and my own install on 12 October 2026, and I have not published a video on Browser Use mode itself yet.

What Hermes Agent Browser Use actually is

There are two questions hiding inside this keyword.

The first is whether Hermes Agent works with Browser Use, the open-source browser automation project at github.com/browser-use/browser-use.

The second is how you get Hermes to use a browser for real work.

I'll take them in that order.

I went looking for a plugin or an MCP connector first, because that is how most tools bolt on to Hermes.

I found that neither exists in the official catalogs, and that neither is needed.

The official Hermes browser docs say Browser Use mode uses the Browser Use CLI 3.0 instead of the built-in browser tools.

They also say the agent writes and executes Python in the browser to click, type, drag and scrape.

Browser Use's own site says the same thing, which is that Hermes Agent ships Browser Use as its default browser driver.

Nous Research makes Hermes Agent, the Browser Use team makes Browser Use, and both projects are MIT licensed.

The quick start

This takes a few minutes on a machine that already has Hermes.

hermes update
hermes tools
hermes

• The first command puts you on a current build, which ships the Browser Use engine.
• The second command opens the tools menu, where you choose Browser Automation and then Local Browser.
• The third command starts a chat.

Then give it a harmless first task.

Open github.com/trending and list the top five repositories with one line on each.

You do not have to switch Browser Use mode on.

The docs say it is the default whenever browser.backend is unset.

You can force it or turn it off in the config file.

# ~/.hermes/config.yaml
browser:
  backend: "browser-use"   # force Browser Use mode
  # backend: "off"         # go back to the built-in browser tools

You can also type /browser use off in a chat to get the older click-by-click tools back.

One version note will save you some head scratching.

My own install is a late-September build of v0.21.4, and its tools menu still lists Browser Use as a row that installs a separate CLI.

The current docs say the engine now ships with every install, so run the update first.

If Hermes says the Browser Use CLI was not found, it has quietly fallen back to the built-in tools.

The three real routes

The driver and the browser are separate choices, and that is the part people mix up.

Browser Use mode is the driver, which is how Hermes controls a page.

The browser is where the page runs, and you can change it without changing the driver.

For Browser Use Cloud, the manual setup is two small edits.

# ~/.hermes/.env
BROWSER_USE_API_KEY=your_key_here

# ~/.hermes/config.yaml
browser:
  cloud_provider: browser-use

The docs add a warning I would not skip.

The key only supplies the credential, and the menu choice decides which browser is active.

Camofox is the one browser that cannot run Browser Use mode, so it keeps the built-in tools automatically.

I explain the whole browser toolset in my Hermes Agent browser guide, and I rank the options in my best browser for Hermes list.

🔥 Want the exact Hermes Agent with Browser Use setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

Where Hermes Agent Browser Use fits in an Agent OS

A browser agent on its own is a demo.

A browser agent with a planner above it, a memory beside it and a checker below it is a worker.

This is how I think about the layers.

Three Hermes features make this easier than it sounds.

• Named sessions let each subagent pass a session name to browser_exec, so every one gets its own browser and they don't fight over a tab.
• Profiles keep things apart, because saved logins are scoped to the Hermes profile that stored them.
• Scheduled jobs can be given the browser toolset, so a morning check can run before you wake up.

There is one rule to remember for bots.

The browser_exec tool is only offered to sessions that have terminal access, because it runs model-written Python on your machine.

A locked-down messaging bot without the terminal toolset keeps the built-in browser tools instead.

Browser use as an activity: three tasks to copy

These are starting prompts, and I am not claiming measured results from them.

Run each one on a low-risk site first.

Research a live page

Open the pricing pages for these three tools and build a table of plan names and monthly prices.
Put the page address next to every number.

The address column is your check.

The docs make an honest point here, which is that web search and web extract are faster and cheaper for plain reading.

Use the browser when a page needs clicks, scrolling or a login.

Fill a form and stop before submit

Go to the contact form on this page and fill it in with the details below.
Do not press submit.
Take a screenshot and show me what you entered.

As far as I can tell from the docs, Hermes does not pause before every click.

The stop instruction is your approval gate, so write it into the task.

Set browser.headed: true while you learn, because it opens a visible window that stays open between turns.

Check something you are logged into

Hermes has two ways to browse as you.

• Real profile browsing, set with browser.use_real_profile: true, drives a managed copy of your active Chrome profile.
• The /browser connect command attaches Hermes to a Chrome window, and it only works in the terminal CLI.

Quit Chrome before a real-profile session, because a running browser can lock the login files.

Open my analytics dashboard and tell me yesterday's total visits.
Do not change any settings and do not click anything that saves, sends or deletes.

If the agent hits a sign-in page with no saved login, Hermes asks you in a masked prompt.

The docs say the password is encrypted on your machine, bound to that site, and filled in without the model ever seeing it.

🔥 Want the exact Hermes Agent with Browser Use setup I run inside my Agent OS?

Inside the AI Profit Boardroom you get the installable Agent OS, step-by-step video tutorials, four coaching calls a week and 3,400+ members building real automations.

→ Get access here

Approvals and safety

Hermes has real guardrails, and it helps to know what each one covers.

The docs are blunt about real profile browsing.

They call it a consent-gated convenience and not an isolation boundary.

A page the agent visits runs with your real logins.

My own rules sit on top of the built-in ones.

• I give the agent a clean profile unless a task truly needs my accounts.
• I keep approvals on smart or manual, and I never run browser jobs in YOLO mode.
• I do the final click on any payment myself.
• I don't ask an agent to get past a CAPTCHA on a site whose terms forbid automation.
• I write "stop and show me" into anything that sends, submits, books or deletes.

What it costs

Browser Use says it has no subscriptions and that purchased credits never expire.

Its README and its pricing page gave different figures for the free signup credit when I checked, so I have left that number out.

The model is usually the larger bill, because a browser job is many decisions in a row.

Honest limits

• The built-in tools read pages as text, so very visual pages can confuse them.
• Page snapshots are cut at 15,000 characters by default, and the full copy is saved to a file.
• The docs list file downloads from the browser as unsupported for the built-in tools.
• The /browser connect command does not run from Telegram, Discord or the web interface.
• Lightpanda has no screenshots, because it has no graphical renderer.
• Browser Use says no browser setup guarantees that every CAPTCHA is avoided or solved.
• Weaker models gain little from Browser Use mode, because they have to write working Python.

That last point comes from Nous Research's own benchmark notes in the Hermes repo.

Their hard task set showed about 60% fewer tokens with Claude Opus 4.8 and about 66% fewer with Kimi K3 at the same success rate.

Their notes also say a small local coder saw no real saving, and that the test used simple practice sites with three or fewer runs per cell.

The MCP route, the standalone CLI and Jev Ultrafast

Browser Use documents two more ways in.

The first is the standalone CLI, which Browser Use's Hermes page installs like this.

uv tool install browser-use
browser-use doctor
browser-use skill install

Hermes then drives the browser-use command through its terminal tool.

The second is MCP.

Browser Use runs a hosted MCP server at https://api.browser-use.com/v3/mcp that takes your key in an x-browser-use-api-key header.

It also has a local one that starts with uvx --from 'browser-use[cli]' browser-use --mcp and needs your own model keys.

Hermes can add a custom server with hermes mcp add, which accepts a --url and a header auth option.

I have not tested that pairing, and Browser Use's MCP page lists Claude, Cursor and Windsurf but not Hermes.

I would only reach for it if you want Browser Use's hosted agent to run a whole task for you.

Jev Ultrafast is a separate project from the same team.

It is a browser agent that lets TypeSafe's Jev model pick every click, and it is not wired into Hermes.

My notes on it are in my Jev Ultrafast GitHub guide.

What to do next

Update Hermes, pick Local Browser, and run the trending-repos test.

Then choose one boring browser job with a clear finish line and add a stop before the final click.

If you want the whole Agent OS around it, with the walkthroughs and four coaching calls a week, it lives inside the AI Profit Boardroom.

FAQ

Does Hermes Agent work with Browser Use?

Yes, and it is built in. The official Hermes docs describe Browser Use mode as the default browser driver, and Browser Use Cloud is a supported cloud browser provider. No plugin or MCP server is needed.

How do I enable Browser Use in Hermes Agent?

Run hermes update, then run hermes tools and open Browser Automation. Pick Local Browser for free local browsing. Browser Use mode is the default driver, so there is nothing else to switch on.

What is browser_exec?

It is the single browser tool the model gets in Browser Use mode. The agent passes it Python code that clicks, types, scrolls and scrapes, so one call can complete a chain of steps.

Is there a Browser Use MCP server for Hermes Agent?

Browser Use publishes a hosted and a local MCP server, but neither is in the Hermes MCP catalog. Hermes can add custom MCP servers with hermes mcp add, although the built-in integration is the simpler route.

How much does Browser Use Cloud cost with Hermes?

Browser Use lists cloud browser sessions at $0.02 per browser-hour and managed residential proxies at $5 per GB, with no subscription. Local browsing in Hermes is free.

Where does Browser Use fit in an Agent OS?

It is the browser hands. A strong model plans the job, Browser Use mode clicks through the page, your notes vault keeps what it found, and a human or second agent checks the result before anything is sent or paid.

Final word on Hermes Agent Browser Use

The integration is already inside Hermes, so your job is to choose the browser, choose a strong model and add your own stop points.

Do that, and you will have a browser worker your Agent OS can rely on through Hermes Agent Browser Use.

Related guides on agentos.guide

→ Jev AI Voice Browser: Talk To Your Browser

→ Hermes Bot Mode

→ Best Hermes Agent Tools

→ The Hermes Command Desk

📺 Video notes + links to the tools 👉

🎥 Learn how I make these videos 👉

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

← all guides on agentos.guide