Browser control ​

Open a browser tab from the center pane's +, or just ask the agent to open a page — it's a real, sandboxed Chromium browser living inside the app, so the agent can fully see and operate it while you watch. Focus the browser with ⌘3 / Ctrl+3 and its URL bar with ⌘L / Ctrl+L.

The agent driving the built-in browser — filling a login form, reading the page and network, and intercepting requests

The agent reuses one browser tab for its work (it navigates that tab rather than piling up new ones). What it can do with a page:

  • Inspect — read the rendered DOM, list network requests, and pull full request/response headers and bodies.
  • Storage — read localStorage, sessionStorage, cookies (including HttpOnly), and IndexedDB — useful for session tokens and auth state.
  • WebSockets — read frames sent and received.
  • Operate — click, type, fill, submit, and scroll to walk through a flow; or run arbitrary JavaScript in the page (approval-gated). The JS console is stateful — variables the agent defines persist across calls — and it can read back recent console output to check results.
  • Screenshot — capture what's rendered; the agent can see the image, and it's saved into your workspace.
  • Intercept traffic — hold requests proxy-style, then continue (optionally modifying URL/method/headers/body) or block them (approval-gated).

Downloads (e.g. opening a .zip) are saved into your workspace so the agent — and you — can open and inspect them.

Safety

Running arbitrary page JavaScript and enabling request interception both pause for your approval — see Approvals & safety.

See also ​

Hacker Sidekick documentation