Set Up Hermes Browser Use Mode Safely

Hermes Browser Use mode gives the agent one browser_exec tool. The agent writes a short Python program that can navigate, click, type, extract data, and recover from simple failures in one run.

On a current Hermes installation, this mode is selected automatically when the Browser Use CLI can run, no browser backend has been forced, and the session has terminal access. If those conditions are not met, Hermes falls back to its built-in browser tools.

Choose the browser source

Browser Use is the driver. You still choose where the browser runs.

Need	Start with	Trade-off
Use an existing local browser	Local Chromium through CDP	Shares the local machine and may expose signed-in sessions
Isolate a test from personal browsing	A separate local browser profile	Requires a clean profile and explicit login
Run remotely or on a schedule	Browser Use Cloud, Browserbase, or another supported cloud browser	Sends page content and actions to that provider
Fetch public pages without a full Chrome session	Firecrawl or Lightpanda, where the task fits	Some interactive sites will not behave like a full browser
Use Camofox	Built-in Hermes browser tools	Camofox has no CDP endpoint for the Browser Use driver

Open hermes tools, select Browser Automation, and configure the provider you intend to use. Adding a provider key does not switch the provider by itself; select it in the tools configuration.

Verify the active mode

Start a fresh Hermes session and run:

/browser status

Confirm the browser source and whether Browser Use mode is active. Then give Hermes a read-only task with a visible success condition:

Open the official Hermes Browser Use documentation.
Return the names of the supported browser providers in a table.
Include the page URL. Do not sign in, download files, or submit forms.

The setup passes if the provider starts, Hermes returns the source URL, and the list matches the page. A confident answer without a URL is not enough.

Force Browser Use or turn it off

To force the driver in ~/.hermes/config.yaml:

browser:
  backend: "browser-use"

To use the built-in browser tools instead:

/browser use off

Or set the backend explicitly:

browser:
  backend: "off"

Restart the session after changing the configuration, then check /browser status again.

Use separate sessions for concurrent browser work

Current Hermes builds support named browser sessions:

browser_exec session=research
browser_exec session=admin

Each name gets separate Browser Use daemon state and, for a cloud provider, a separate cloud browser. Separation prevents two agents from clicking in the same tab. It does not make sensitive work safe by itself; each session still inherits the permissions and credentials of its browser source.

Compare it with the built-in tools

Run the same bounded task twice, once with Browser Use enabled and once with it off. Record:

• whether the result is correct;
• how many retries were needed;
• whether the source URL and relevant evidence were returned;
• whether a recording or audit log exists;
• the model and provider cost shown by your setup.

Use the mode that is more reliable for your actual sites. Vendor efficiency claims are useful hypotheses, not proof for your workload.

Troubleshooting

Problem	Check
Hermes shows only built-in browser tools	Confirm the Browser Use CLI is runnable and the session has terminal access.
The wrong cloud provider opens	Select the provider in hermes tools; a saved API key alone does not select it.
A local browser will not connect	Start a supported Chromium browser with remote debugging, or use Hermes’ connect flow.
Two tasks interfere with each other	Give each task a different session= name.
A private URL is blocked	Do not weaken the boundary by default. Decide whether that domain and browser profile are safe first.
The task asks for approval-sensitive action	Stop before submission and hand control to the user.

Official sources

• Hermes Browser Use mode
• Hermes browser provider configuration
• Browser Use CLI