Actions Overview¶

strands-adb exposes one tool: adb. It dispatches on a single action="..." argument. There are 90+ actions grouped into domains.

from strands_adb import adb

adb(action="screenshot")
adb(action="smart_tap", text="Send")
adb(action="battery")

Action Groups¶

graph TD
    ADB["🔧 adb(action=...)"] --> DEVICE["📱 Device"]
    ADB --> UI["🎯 UI"]
    ADB --> SCREEN["👁️ Screen"]
    ADB --> APPS["📦 Apps"]
    ADB --> FILES["📂 Files"]
    ADB --> INTENT["🔗 Intents"]
    ADB --> SENSOR["🌡️ Sensors"]
    ADB --> SETTINGS["⚙️ Settings"]
    ADB --> LOGS["📜 Logs"]
    ADB --> A11Y["♿ Accessibility"]
    ADB --> CAMERA["📸 Camera"]
    ADB --> COMMS["📞 Comms"]

    style ADB fill:#3DDC84,color:#000

📱 Device¶

Action	Purpose
`list_devices`	Enumerate connected devices
`select_device`	Set default device serial
`device_info`	Model, manufacturer, Android version
`battery`	`dumpsys battery`
`wake`	Wake screen
`unlock`	Wake + swipe + optional PIN

🎯 UI Input¶

Action	Purpose
`tap`	Tap at `(x, y)`
`swipe`	Swipe between two points over duration
`type_text`	Type text into focused field
`key`	Send keyevent (back, home, enter, volume_up, ...)
`back` / `home` / `recent`	Convenience keyevents
`gesture_long_press`	Long press at `(x, y)`
`gesture_path`	Multi-point gesture sequence
`gesture_pinch`	Pinch-to-zoom
`gesture_stream`	Stream of gestures

👁️ Screen & UI Inspection¶

Action	Purpose
`screenshot`	Capture PNG + return as Converse image block
`screen_record`	Blocking fixed-duration recording
`screen_record_start` / `_stop` / `_status`	Background recording — record while agent acts
`screen_frames`	Extract N live screenshots at intervals
`video_frames`	Extract frames from an on-device video
`ui_dump`	Raw UIAutomator XML
`ui_find`	Find UI node by text / resource-id / content-desc
`ui_tap_by`	Tap a node found by `ui_find` criteria
`ui_wait_for`	Poll until a node appears
`smart_tap`	High-level: tap by semantic text

→ Vision guide · Smart tap guide · UI automation

📦 Apps¶

Action	Purpose
`list_packages`	List installed apps
`launch`	`am start` a package
`kill`	Force-stop a package
`install`	`adb install` an APK
`uninstall`	Uninstall a package
`clear_data`	Clear app data
`current_app`	What's in the foreground

→ Apps & Intents guide

📂 Files¶

Action	Purpose
`push`	Host → device
`pull`	Device → host
`ls`	List remote directory

🔗 Intents¶

Action	Purpose
`open_url`	`am start -a VIEW -d URL`
`share_text`	Android share sheet
`start_activity`	Arbitrary `am start` with args

🌡️ Sensors & Thermals¶

Action	Purpose
`sensors`	Accelerometer, gyro, light, etc.
`thermals`	CPU / skin / battery temps
`wifi_info`	SSID, BSSID, signal

→ Sensors guide

⚙️ Settings Mutation¶

Action	Purpose
`setting_get` / `setting_put` / `setting_delete` / `setting_list`	Raw namespace-scoped settings
`set_ringer`	silent / vibrate / normal
`set_brightness`	0–255
`set_bluetooth`	enable/disable
`set_airplane_mode`	enable/disable
`setting_dump`	Dump all settings in a namespace

→ Settings guide

📜 Logs¶

Action	Purpose
`logcat`	One-shot `logcat -d`
`log_stream_start` / `stop` / `status`	Background streaming → DevDuck event bus
`notifications` / `notifications_parsed`	Dump notifications (raw / structured)

→ Logcat streaming guide

♿ Accessibility¶

Action	Purpose
`accessibility_list`	List available services
`accessibility_toggle_service`	Enable/disable a service
`accessibility_system_action`	Trigger a system action
`accessibility_captions`	Toggle live captions
`accessibility_magnification`	Enable screen magnification
`accessibility_font_scale`	Set font size multiplier
`accessibility_status`	Overall a11y state

→ Accessibility guide

📸 Camera¶

Action	Purpose
`camera_photo`	Take a still photo (returns image block)
`camera_video`	Record a video

→ Camera guide

📞 Comms¶

Action	Purpose
`dial`	Open dialer with number (or auto-call)
`sms_compose`	Draft an SMS
`media_control`	play/pause/next/prev/stop
`volume`	Raise/lower/mute

Response Shape¶

Every action returns:

{
    "status": "success" | "error",
    "content": [
        {"text": "human-readable summary"},
        # optionally, for screenshots / camera:
        {"image": {"format": "png", "source": {"bytes": b"..."}}},
    ],
    # action-specific extras:
    "path": "/tmp/foo.png",
    "devices": [...],
    "info": {...},
}

The content field follows the AWS Converse API content block format, which Strands Agent consumes directly.

What's Next¶

Vision — screenshots as image blocks
Smart Tap — semantic UI automation
DevDuck — production agent runtime
API Reference — full action signatures