List connected adb devices (emulators and physical) and available AVDs.

Start an Android emulator (AVD) and wait until it has booted. Returns immediately if it is already running.

Screen size, density, coordinate scale and the app in the foreground.

Lock the screen to portrait or landscape, or re-enable auto-rotation.

List the apps this server may use. Only debug builds installed from Android Studio (Run) are allowed: debuggable + testOnly + installed via adb.

Package and activity currently in the foreground.

Launch an allowed dev app by package (its launcher activity), or a specific activity of it.

Force-stop an app.

Delete all data of an app (like a fresh install). Also stops it.

Open a URL / deep link in an allowed dev app with an ACTION_VIEW intent.

Take a screenshot (costs far more tokens than get_ui; use it for visuals the tree lacks). Its pixel coordinates are the screen coordinates every other tool uses. With region, returns a sharper crop for reading small text; a crop's pixels are NOT screen coordinates.

The visible UI as a list of elements: [ref] (stable across calls), text, desc, id, centre, size, flags. Target elements by ref (index), text or id. Falls back to a screenshot when no tree is available.

Wait until an element with the given text/id appears (or, with gone=true, disappears).

Read the log of an allowed dev app (including lines of crashed earlier processes). By default returns only lines that are new since the previous call with the same filters.

List recorded crashes, ANRs and native crashes of a dev app, newest first, with exception and top frame. Reports survive the process and logcat rotation.

Full report of one crash from list_crashes: header (process, version, time) and the complete stack trace.

Tap an element (by index, text or id) or a screen point (x, y).

Long-press an element (by index, text or id) or a screen point (x, y).

Swipe/drag from one screen point to another. A slow swipe (duration >= 1000) acts like a drag.

Read a whole scrollable list or form in one call: scrolls it to the end and returns every entry it contains, in order. Much cheaper than scrolling step by step. Afterwards the list stays at the end.

Scroll the content in a direction (down = reveal content below). Scrolls the given element (index), otherwise the largest scrollable element on screen (the screen centre only when no UI tree is available).

Type text into the focused field. Optionally tap a target first to focus it, clear it, and press enter afterwards.

Press a key: back, home, enter, del, backspace, forward_del, tab, escape, app_switch, menu, search, power, wakeup, volume_up, volume_down, dpad_up, dpad_down, dpad_left, dpad_right, dpad_center, move_home, move_end, any KEYCODE_* name, or a key code number.