Features

Watch and take over a Hermes Agent bot’s screen

On a headless Linux host, Hermes Agent can give each bot its own desktop for computer use and a visible browser. Herald shows that Bot Screen live on your phone and lets you take over when the bot hits a login, a 2FA prompt or a CAPTCHA, then hand control back so it carries on. Hosts running macOS or Windows don’t offer it.

At a glance

Hermes feature
Bot Screen: an Xfce desktop per bot on a headless Linux gateway host
Hosts
A Linux server, a cloud VM, Hermes Cloud or the official Docker image
Not on
macOS or Windows hosts, which have a real display
Needs on the host
TigerVNC and Xfce packages, and Computer Use enabled for the bot
In Herald
A live view with Take over and Hand back

Where to find it

The Bot Screen card sits under the pet. On iPhone, tap the pet in the chat header; on iPad, look in the inspector rail. While the bot drives its screen, the card shows a thumbnail that refreshes every few seconds and says who is in control: Bot in control, You have control or Someone else has control.

Tap the card for the live view. It fills the screen, and dragging across the desktop won’t close it by accident.

Take over and hand back

  1. Watch first

    The view streams the bot’s desktop. Fit to view shows all of it; Actual size shows it pixel for pixel, and a one-finger drag pans.

  2. Tap Take over

    Your taps and typing now drive the screen, and the bot waits until you hand back. Sign in, enter the code or solve the CAPTCHA.

  3. Type what you need

    The Keyboard button opens the keyboard. Esc and Tab are in the toolbar, and a hardware keyboard works too.

  4. Tap Hand back

    The bot gets its screen back. Tell it in the chat that it can carry on.

If you see a different message

Bot Screen messages in Herald and what to do
MessageWhat to do
Screen is offTap Start screen
Screen not installedRun the command Herald shows, once, on the host. Herald never runs sudo for you.
Someone else took over. You are watching.Another device holds control. Take over again or wait.
Someone else has control, from a session you leftTap Hand back anyway to clear the old takeover
The screen stopped streaming.Tap Reconnect

Good to know

Leaving Herald keeps control, so you can switch apps to fetch a 2FA code and come back to the same session. Closing the live view hands control back.

Hermes stops a screen nobody uses after 30 minutes by default and starts it again on next use, and it checks for free memory before starting one. The Hermes docs suggest planning about 1.1 to 1.5 GB per open screen with a browser.

Limits and security

  • Linux hosts only.
  • Herald doesn’t install packages. On the official Docker image, an operator runs the install line once as root in the container.
  • A screen is a work surface, not a security boundary: bots on one host share its user account, files and network.
  • While you hold control, the bot’s computer-use and browser tools are refused, but that is a fence in the tools, not the operating system. Don’t type secrets into a bot you wouldn’t trust with them.

Questions

Why don’t I see a Bot Screen card?

Either the host runs macOS or Windows, or its Hermes doesn’t have Bot Screen yet. In both cases Herald hides the card rather than show an error.

Does the bot keep working when I close Herald?

Yes. The screen lives on the host, not on your phone, so the bot carries on.

Can two people watch the same screen?

Yes. The most recent Take over wins, and the previous controller drops back to watching.

Do I need Hermes Cloud?

No. Any headless Linux host works, including your own VPS or the official Docker image.

Sources

Get Herald for Hermes Agent

iPhone, iPad, Apple Watch, CarPlay and Android. $12.99 once, no subscription, no account and no Herald server.

Herald is an independent, unofficial app. It is not made, endorsed or supported by Nous Research.