Start a scrcpy session for fast input control and screenshots. When a session is active, tap/swipe/text/screenshot are 10-50x faster. Requires scrcpy-server to be installed.

Stop the active scrcpy session. Tools will fall back to ADB commands.

Report which scrcpy version is being used by the MCP server. The version is detected from the SCRCPY_SERVER_VERSION environment variable, the scrcpy --version binary, or a built-in default.

List all connected Android devices with their serial numbers, state, and model

Get detailed info about a device: model, Android version, screen size, SDK level, battery level

Wake the device screen (turn screen on)

Turn the device screen off

Enable WiFi ADB and connect to the device wirelessly. Returns the connection address.

Disconnect from a wireless ADB device

Rotate the device screen (requires active scrcpy session)

Expand the notification panel (requires active scrcpy session)

Expand the quick settings panel (requires active scrcpy session)

Collapse all open panels (notification, settings) (requires active scrcpy session)

Take a screenshot of the Android device screen. Returns the image as base64.

Start recording the screen. Recording continues until screen_record_stop is called.

Stop screen recording and optionally pull the file to the host.

Tap at the specified screen coordinates

Perform a swipe gesture from one point to another

Perform a long press at the specified coordinates

Perform a drag and drop gesture from one point to another. Uses input draganddrop on Android 8.0+ (API 26), falls back to swipe on older versions.

Type text into the input field. Reports device context (screen, app, session). When submit=true, auto-detects and taps the send button. Optionally target a specific input field by text label, resource ID, or content description instead of auto-focus.

Send a key event to the device. Supports keycodes like HOME, BACK, ENTER, VOLUME_UP, etc.

Scroll at the specified position. dy=negative scrolls UP (reveals content above), dy=positive scrolls DOWN (reveals content below). Use large values (e.g. -400, 500) for meaningful scroll distance.

Launch an app by package name. PREFERRED over swiping the app drawer / manual UI navigation. Uses scrcpy START_APP (fast) when session is active, falls back to ADB am start. Supports force-stop prefix (+).

Force-stop an app on the device.

Install an APK file on the device from the host machine.

Uninstall an app from the device.

List installed packages on the device. Optionally filter by name or restrict to system/third-party apps. Use to find package names for app_start.

Get the currently foregrounded app package name and activity.

Get the current clipboard content from the device. Uses scrcpy GET_CLIPBOARD when a session is active (works on Android 10+), falls back to ADB clipboard commands.

Set the clipboard content on the device. Uses scrcpy SET_CLIPBOARD when a session is active (with optional paste flag), falls back to ADB clipboard commands.

Dump the full UI hierarchy of the current screen as XML.

Find UI elements on screen by text, resource ID, class name, or content description. Returns matching elements with their tap coordinates.

Find a UI element by text, resource ID, class name, or content description, and immediately tap it. Smartly finds the nearest input field when targeting a label.

Dump the UI and return a clean Markdown-like tree of visible and clickable elements with viewport coordinates. NOTE: WebView elements may report identical y-coordinates when scrolled; use scroll_to_element first to bring target fields into view.

Wait for a UI element to appear on screen. Polls the UI until the element matches criteria or timeout is reached.

Find a UI element, tap it, inject text, and optionally press ENTER. Smartly finds the nearest input field when targeting a label.

Wait for a specified number of milliseconds. Use between actions to wait for animations, page loads, or network responses.

Scroll the page to bring a specific UI element into view. Tries to find the element by text, then swipes up/down until it becomes visible or timeout is reached.

Fill a multi-field form. Provide an array of {label, value} pairs. For each pair, finds the nearest input field to the label text, taps it, and types the value. Handles checkboxes and dropdowns by label matching.

Execute an arbitrary ADB shell command on the device and return the output. Use this for any device operation not covered by other tools.

Push a file from the host machine to the device.

Pull a file from the device to the host machine.

List directory contents on the device.

Start an HTTP MJPEG video stream of the device screen and open a viewer window (ffplay). Starts a scrcpy session automatically if needed.

Stop the HTTP MJPEG video stream and close the viewer window for a device.