Side panel, browser, and desktop apps

Press Cmd+J (Ctrl+J on Windows and Linux) to open the panel beside the chat. It has three tabs: Terminal, Files, and Browser.

Terminal

The terminal opens in the agent’s workspace folder. It is a real terminal, useful for running a script the agent wrote or checking a file path. Shell commands the agent runs still pass the approval gate; commands you type yourself are your own.

Files

The Files tab browses the workspace:

preview documents and images;
attach a file to the next message;
open an .html file and press Run to view it in the sandbox;
see the files the agent created or changed.

Browser

The Browser tab is a real browser with separate sign-in profiles:

Keep a work profile apart from a personal one.
Import cookies from Chrome, Edge, Brave, Arc, Chromium, or Firefox to stay signed in.
Let the agent fill in forms, check dashboards, and collect information in the panel.

Looking at a page runs freely. Clicking and typing ask first. Pages on private networks are never opened by the agent’s web tools.

Desktop apps

On macOS and Windows, Settings › Computer use lets the agent use desktop apps: read a screen, click, and type.

Looking runs freely.
Clicking and typing ask first.
PP Agent itself, terminals, and password settings are always off limits.
On macOS, the operating system asks for accessibility and screen-recording grants.

This is for apps without an API or MCP server: a legacy ERP, a desktop reporting tool, or an internal application.

Which one to use

Task	Use
Run a script or inspect a path	Terminal
Preview a generated file	Files
Sign in to a web app and check something	Browser
Operate an app with no API	Computer use
Reach a business system with an API	MCP server