Browser Node
A real Chrome that keeps your logins.
Overview
The Browser node gives a workflow or an agent a real Chrome browser. Attached to an Agent, the agent decides what to open, click and type. When a site needs a login, the agent asks the user to take over in the live browser view, and the browser keeps that session for next time. Used as a step, the node opens a page or reads the page that is currently open.
Configuration
| Field | Description | Default |
|---|---|---|
Action |
Open a page or Read the current page |
Open a page |
URL |
The page to open when the node runs as a step. Supports {{variables}}. Only for Open a page |
- |
Allowed sites |
Comma-separated domains. linkedin.com also covers www.linkedin.com. The browser can't open other sites while this node runs. Leave empty to use the browser's own list |
- |
On Error |
Stop Flow, Continue (empty output) or Continue (default value) |
Stop Flow |
Output Variable |
Variable that holds the page | page |
Example: Agent that works on a site you log in to
Chat Start → Agent (Browser attached, Allowed sites: linkedin.com) → Send Response
The first time, the agent asks you to log in through the live view. After that it can open, read and fill in pages on that site for you.
Best Practices
- Always set Allowed sites to the few domains the task needs.
- Prefer an API (HTTP Request or an MCP app) when the site offers one; the browser is for sites that don't.