Desktop & Browser Agent

Get it for macOS

Lets your agents use the apps and browser tabs on your machine in the background, while you keep working. Agents read and operate native apps and Chrome, Edge or Brave windows through Cua Driver from Cua AI, which the settings page installs and checks on each machine. The optional Browser Bridge for bb extension lets them work in any tab of your signed-in browser, in an Agent tab group you can pause. Tools that run code or force-quit apps start off, and agents can be made to ask before clipboard, screen capture and file actions. Needs Cua Driver (free, MIT). Tested on macOS 14+; Windows 10/11 and Linux with X11 are supported by Cua Driver but not yet tested with this plugin.


Overview

View source

What you get

Agents in any bb thread can open, read and operate the apps on the machine that runs the thread: native windows, Electron apps and Chrome, Edge or Brave windows. They act in the background through the accessibility tree, so your pointer and keyboard stay yours, and they check the result before they report.

With the optional Browser Bridge for bb extension, agents also work in the tabs of the browser you are signed in to: Brave, Chrome or Edge. Tabs they open go into a purple Agent group, a cursor shows where they click, and Pause agents in the toolbar stops them at once. They fill whole forms in one step, answer page dialogs and attach files without a file picker.

You decide what agents may do

  • Tools that run code, force-quit apps, record the screen or change the driver start off. Every tool can be turned on, off or kept out of the prompt until an agent asks for it.
  • Ask before sensitive actions makes agents wait for your answer in the thread before they read the clipboard, capture the whole screen, upload or download files, attach to a signed-in browser or first use your browser tabs.
  • Providers with their own computer use, such as Codex, keep it by default. You can give every provider these tools instead, set one provider apart, or turn them off.
  • The agent skill tells agents to prepare a send, submit, delete or purchase and ask before they do it, and to leave CAPTCHAs and permission prompts to you.

Setup

The settings page shows every machine that runs your threads, with its driver version, service state and macOS permissions. Install Cua Driver… installs the release this plugin was tested with, verified by checksum, and Grant macOS permissions walks you through Accessibility and Screen Recording. Remote machines work the same way. When Cua Driver adds tools, agents get them without a plugin update, and the settings page lists what changed.

Everything is also available from the terminal with bb cua.

Requirements

  • bb 0.39 or later.
  • Cua Driver from Cua AI, open source under the MIT license, on each machine agents should drive. Nothing installs until you click install. It sends content-free telemetry to Cua by default; cua-driver telemetry disable turns it off.
  • Tested on macOS 14 or later. Windows 10/11 and Linux with X11 are supported by Cua Driver but not yet tested with this plugin.
  • The browser extension is optional and comes from the Chrome Web Store. It is tested on macOS; its Linux setup is untested, and it does not run on Windows.

A community plugin, not affiliated with Cua AI or the bb team. Source, reference and changelog: https://github.com/qwadrox/bb-plugin-desktop-browser-agent

More in Agents & Providers 62

Add, choose, configure, route, or coordinate who runs a thread.

View all