Skip to content

Desktop & Browser Automation

Silicon Worker can operate your desktop apps and web pages directly — not by guessing coordinates from screenshots, but by reading the real elements on screen and clicking, typing, and so on. This lets it handle work that has no API and would otherwise require manual clicking.

What it can do

  • Desktop automation: operate desktop apps on your machine, including reading the current UI elements, clicking, typing text, pressing keys, scrolling, and waiting. Useful for getting work done in apps that have no API.
  • Browser automation: automate web pages, including opening a URL, reading page content, clicking, filling forms, selecting dropdown options, scrolling, going back, and extracting information from the page.

Have it operate a desktop app

  1. Make sure you've granted the system permissions needed for desktop automation (see below).
  2. Describe the task clearly in the chat, naming which app to act in and what to do.
  3. It reads the current UI, identifies the actionable elements, and performs clicks, typing, and other actions step by step. The process is visible to you.

Have it automate a web page

  1. Describe the web task in the chat — for example, open a URL, fill and submit a form, or extract certain information from a page.
  2. It opens the page, reads its structure, then clicks, fills forms, pages through, and extracts content as needed.

Login state

The browser keeps your login state, so for sites that require sign-in, once you've logged in once, later tasks can usually reuse that state without signing in again each time.

Permissions you need

Desktop automation requires system-level accessibility/automation permissions:

  1. On first use, Silicon Worker prompts you to authorize and guides you to System Settings to enable the relevant permission.
  2. Complete the authorization as prompted before continuing.

For details on permissions, see Permissions and the Permissions reference.

Things to keep in mind

Key actions are confirmed first

Sensitive actions such as submitting forms or deleting are recognized as high-risk; Silicon Worker pauses and asks for your consent before executing them.

  • During automation the UI is actually operated, so try it first on a non-critical window or a test page.
  • You can stop an action in progress at any time with the stop button.

Result and verification

The action panel shows what it did at each step; when the task finishes, check the relevant app or page to confirm the result matches your expectation.