Run your first agent task

After installing the app, use the bundled runtime to open a page and read the text and controls an agent can act on.

Read the installed guide

beaconctl skill show

This prints the guide that matches the installed browser. Use it for helper names and current engine limitations.

Open a page and inspect it

Run this complete script in your terminal:

beaconctl nodejs <<'JS'
await useOrCreateTaskSpace('first-browser-task');
await openOrReuseTab('https://example.com', { wait: true });
cliLog(await snapshotText());
await completeTaskSpace();
JS

A marked agent tab opens Example Domain. The terminal prints the page's semantic text and link. Completion keeps the tab for your review and ends agent control of the task.

Helpers are available without imports. cliLog writes results to the terminal; script return values are not printed automatically.

Continue working

Declare the task in every script. Reusing its name resumes its working tab during the current browser session, and openOrReuseTab navigates that tab. Completion leaves the browser idle; declare the task again to continue. After Take Over, the person can choose Resume Agent to continue agent control, or dismiss the takeover message so a newly declared task can start. Both require an explicit human action.

Closing the Node process does not close tabs or finish a task automatically. Browser controls explains how to take over, stop work, and handle restarts. For page actions and helper discovery, read agent scripts.