Agent Browser logo

App Connector · Browser

Agent Browser

A token-light browser the assistant can drive.

Drive a browser with compact accessibility snapshots that spend fewer tokens.

What you can do

Turn the connector on with your own account and the assistant works inside Agent Browser.

Navigate

Open and move through web pages on your behalf.

Interact

Read page contents, click elements, and fill forms.

Reuse your logins

Point it at a Chrome profile for a read-only copy of your signed-in sessions, or a folder to keep a profile across runs.

Spend fewer tokens

Reads pages as compact accessibility snapshots with element refs instead of full HTML, so routine automation stays cheap.

Things you can ask

Type these in chat once the connector is on.

Open the dashboard and read out the weekly numbersFill the demo signup form with test dataCheck whether the pricing page renders the new plan

How you sign in

No account. The first run downloads Chrome for Testing (about 150 MB); later runs reuse it. Pick a Chrome profile, headless or visible, proxy, downloads, and other settings at install or later from the tools sidebar.

Activation is per workspace, so the connector can be on for one project and off for another.

Four browser connectors

This is the one that runs unattended. It opens its own window with nothing to load into a browser and nothing to click, so it works inside scheduled workflows and while you are away, and its compact snapshots keep long sessions cheap. It starts signed out until you point it at a Chrome profile. Browser MCP attaches to the browser you are already signed into. Chrome MCP goes further into Chrome, at a higher setup cost. Playwright is the one for end-to-end tests. The install screen names the difference and switches between them in one click.

Connect Agent Browser to Plexon

Install Plexon, open the Apps view, and turn on Agent Browser. From then on you just ask.