Run a Loop on Your Gemini Subscription

In this tutorial, you will configure Google Gemini Web as an automated browser worker in Zenborg. You will log in once via the dedicated Chrome profile, configure a loop, and execute tasks without paying per-token API fees.

What You Need Before You Start

Zenborg installed on macOS or Windows.
Google Chrome or Chromium installed on your system.
An active Google account with access to Gemini.

Ordered Screens Overview

During this tutorial, you will navigate through four screens in sequence:

Screen 1: Browser Settings Panel — Launching the isolated browser profile.
Screen 2: Dedicated Chrome Profile — Authenticating with Google.
Screen 3: Loop Orchestration Form — Selecting gemini-web as Worker.
Screen 4: Execution Dashboard — Watching automated browser stages run.

Step-by-Step Instructions

Step 1: Open the Dedicated Browser Profile

In Zenborg, open the Loops tab or navigate to Settings > Browser Workers.
Locate the button labeled “Open browser to log in”.
Click the button to launch the automated browser instance.

A clean, dedicated Google Chrome window opens.
Notice that this profile (~/.zenborg-browser) does not share your personal browsing data.
Log in to your Google account with your standard credentials.
Navigate to https://gemini.google.com/app and confirm the chat interface loads.
You can now leave the browser window open or minimize it.

Step 3: Configure a Loop Using Gemini Web

Return to the Zenborg Loops tab.

In the task description area (Task Scope), enter your feature request:

Refactor the authentication middleware to support JWT refresh token rotation.

Set Manager Agent to antigravity (or claude).
Set Worker Agent to browser with gemini.
Set Smart worker (optional) to another CLI agent you have (such as hermes), or leave as None (default).

Step 4: Start the Loop

Click “Start Orchestration Loop”.
Watch the stage timeline activate.

Step 5: Understanding Browser Execution and Notices

During the execution stage, Zenborg sends structured prompts to Gemini Web through Chrome DevTools Protocol.
If Gemini Web presents a CAPTCHA and you have configured a fallback agent, the loop continues automatically and logs: "Google Gemini hit a CAPTCHA; the loop continued on another agent."
If no fallback is active, Zenborg shows a login banner with buttons to "Open browser to log in" and "Resume loop".

[Visual]	[Spoken]
00:00 - 00:15 Opening Zenborg. Mouse moves to the Browser Worker section and clicks `"Open browser to log in"`.	“Did you know you can use your standard Google Gemini subscription in Zenborg without paying for API tokens? Let’s set up a Browser Worker in under two minutes.”
00:15 - 00:30 Chrome window launches with dedicated profile flag. User enters credentials into Google login and reaches the Gemini app interface.	“Zenborg launches a dedicated, sandboxed Chrome profile. You sign in to your Google account just once. Your session cookies stay safely stored on your machine.”
00:30 - 00:48 User returns to Zenborg. In the Loop configuration, selects Antigravity or Claude Code as Manager and Browser Gemini as Worker. Enters task scope.	“Now in Zenborg’s loop builder, we choose our Manager Agent to draft our plan, and select Browser Gemini as our worker. We can also configure a Smart Worker fallback in case Gemini hits a rate limit.”
00:48 - 01:05 User clicks `"Start Orchestration Loop"`. In split-screen, the timeline shows prompt dispatching while Chrome streams Gemini responses.	“Click Start Orchestration Loop. Zenborg sends prompts to Gemini Web, extracts the code diffs, and applies changes directly inside an isolated Git worktree.”
01:05 - 01:20 Execution finishes. Loop diff viewer shows clean refactor with zero API token charges.	“The feature is implemented, tests pass, and your API bill remains at zero dollars. That’s the power of Zenborg Browser Workers.”