Commands

Commands are Playwright-like browser instructions sent from the worker to the client.

How commands work

Commands are relayed one at a time through the session:

1. Worker sends worker.command
2. Session actor relays as server.command to the client
3. Client executes the command in the webview
4. Client sends client.command_response
5. Session actor relays as server.command_result to the worker

Command types

Commands mirror Playwright’s browser automation API:

Navigation

• navigate — Navigate to a URL
• reload — Reload the current page
• goBack — Navigate back
• waitForNavigation — Wait for a navigation to a URL matching a pattern

DOM interaction

• click — Click an element by selector
• fill — Fill an input field
• select — Select a dropdown option
• waitForSelector — Wait for a DOM element to appear

DOM inspection

• getAttribute — Get an element attribute value
• textContent — Get an element’s text content
• innerHTML — Get an element’s inner HTML
• isVisible — Check if an element is visible

JavaScript

• evaluate — Execute JavaScript in the page context

Page state

• url — Get the current URL
• title — Get the page title
• content — Get the full page HTML
• screenshot — Capture a screenshot

Input

• press — Press a keyboard key

Waiting

• waitForNetworkIdle — Wait for network activity to settle
• waitForTimeout — Wait for a fixed duration

Status UI

• updateStatus — Update the overlay heading, subheading, or progress

One at a time

Only one command can be pending at a time. The worker must wait for the result before sending the next command.

Next steps

• Yield to User — Hand control to the user
• Protocol: Commands — Wire format details