Quickstart¶
From zero to a phone-controlling agent in under 2 minutes.
The Journey¶
graph LR
A["1️⃣ Install"] --> B["2️⃣ Connect"]
B --> C["3️⃣ Import adb"]
C --> D["4️⃣ Ask anything"]
style A fill:#1e3a5f,stroke:#60a5fa,color:#fff
style B fill:#3DDC84,stroke:#3DDC84,color:#000
style C fill:#4a1d96,stroke:#a78bfa,color:#fff
style D fill:#92400e,stroke:#fbbf24,color:#fff
1. Install¶
2. Connect¶
If you see unauthorized, accept the trust dialog on the phone.
3. Hello, Phone¶
from strands import Agent
from strands_adb import adb
agent = Agent(tools=[adb])
agent("what's on my phone screen right now?")
The agent will:
- Call
adb(action="screenshot") - Receive a PNG image block (it can literally see)
- Describe what's on screen
4. Drive the UI¶
Under the hood:
sequenceDiagram
participant U as 🧑 You
participant A as 🤖 Agent
participant T as 🔧 adb tool
participant P as 📱 Phone
U->>A: "open whatsapp & read last msg"
A->>T: action="launch", package="com.whatsapp"
T->>P: am start com.whatsapp
P-->>T: launched
A->>T: action="screenshot"
T->>P: screencap -p
P-->>T: PNG bytes
T-->>A: image block
A->>A: (vision) read chat
A-->>U: "Mom: on my way back, 20m"
5. Common One-Liners¶
# Reality check
agent("take a photo with the front camera and describe me")
# Device state
agent("how's my battery? any thermal warnings?")
# Notifications
agent("read me my current notifications")
# Launch app
agent("open spotify and tell me what's playing")
# UI automation
agent("tap the button that says 'Send'")
# Settings mutation
agent("enable airplane mode, I'm about to board")
# Sensors
agent("is the phone face-down right now?")
6. DevDuck (Recommended)¶
DevDuck is the minimalist agent runtime. With one env var, adb becomes a first-class tool:
export DEVDUCK_TOOLS="strands_adb:adb;strands_tools:shell"
devduck "text mom 'on my way' via whatsapp"
7. Multi-Device¶
Target a specific device when you have several connected:
agent("take a screenshot of device 59230DLCH0012Z")
# Or set a default:
import os
os.environ["ADB_SERIAL"] = "59230DLCH0012Z"
→ See Connect a Device for wireless / SSH / multi-device setups.
What's Next¶
graph LR
QS["✅ You are here:<br/>Quickstart"] --> V["👁️ Vision"]
QS --> SM["🎯 Smart Tap"]
QS --> DD["🦆 DevDuck"]
QS --> EX["📚 Examples"]
style QS fill:#3DDC84,stroke:#3DDC84,color:#000
style V fill:#1e3a5f,stroke:#60a5fa,color:#fff
style SM fill:#4a1d96,stroke:#a78bfa,color:#fff
style DD fill:#92400e,stroke:#fbbf24,color:#fff
style EX fill:#831843,stroke:#f472b6,color:#fff
- Vision / Screenshots — how the agent actually sees
- Smart Tap — semantic UI automation
- DevDuck Integration — 1-line agent runtime
- Examples — WhatsApp, notifications, autonomous