DroidAI DroidAI

Select Language

🇺🇸 English
🇰🇷 한국어
🇯🇵 日本語
🇨🇳 中文
🇪🇸 Español
🇫🇷 Français
🇩🇪 Deutsch
🇧🇷 Português
🇻🇳 Tiếng Việt
🇹🇭 ไทย
DroidAI Logo

DroidAI

AI-Powered Android Device Controller
Complete User Guide & Tutorial

Control one phone or an entire farm of devices using natural language. Just tell the AI what to do — it sees the screen, taps, types, and scrolls for you.

Version 1.0 • March 2026

Table of Contents

1. Introduction & System Requirements

What is DroidAI?

DroidAI is a desktop application that lets you control Android devices using natural language. Instead of manually tapping through apps, you simply describe what you want done — and DroidAI's AI agent takes over. It observes the device screen, decides which actions to take, executes them, and repeats until the task is complete.

DroidAI supports controlling multiple devices simultaneously, making it ideal for phone farms, testing labs, and automation workflows at scale.

How It Works

You type a command → AI reads the device screen (UI tree + screenshot)

AI decides an action (tap, type, scroll, etc.) → Executes via ADB

AI observes the result → Decides next action → Repeats until done

System Requirements

ComponentRequirement
Operating SystemWindows 10/11 (64-bit)
RAM4 GB minimum, 8 GB recommended
GPUOpenGL ES 2.0 compatible (for device mirroring)
Android DevicesAndroid 9 (API 28)+
USBUSB 2.0+ cable (data-capable, not charge-only)
InternetRequired for LLM API calls
LLM API KeyAnthropic, OpenAI, Gemini, DeepSeek, Grok, or custom

2. Installation & First-Time Setup

Step 1: Install DroidAI

  1. Download the latest installer from droidai.live
  2. Run the installer and follow on-screen instructions
  3. Launch DroidAI from the Start Menu or desktop shortcut
  4. Sign in with your Google account when the login screen appears

Step 2: Install ADB (Android Debug Bridge)

ADB is required for DroidAI to communicate with your Android devices. DroidAI bundles its own ADB, but if you encounter issues:

  1. Download Android SDK Platform Tools from developer.android.com
  2. Extract to a folder (e.g., C:\platform-tools)
  3. Add the folder to your system PATH environment variable
  4. Open a terminal and verify: adb version

Step 3: Enable USB Debugging on Your Phone

  1. Open Settings on your Android device
  2. Go to About Phone (or Software Information)
  3. Tap Build Number 7 times until you see "You are now a developer"
  4. Go back to Settings → Developer Options
  5. Enable USB Debugging
  6. Connect your phone to the PC via USB cable
  7. Tap Allow when the USB debugging authorization prompt appears. Check "Always allow from this computer".
Use a data-capable USB cable. Many cables that come with chargers are charge-only and will not work with ADB.
For Samsung devices, you may also need to enable "USB debugging (Security settings)" in Developer Options.

Step 4: Verify Connection

adb devices

You should see your device serial number followed by device:

List of devices attached
ABCD1234    device

If you see unauthorized, check your phone for the authorization popup.

3. Connecting Your First Device

USB Connection

  1. Connect your phone via USB with debugging enabled
  2. Launch DroidAI — your device should appear automatically in the device grid
  3. You'll see a live screen mirror of your device
  4. The device panel shows the device name and a numbered badge (#1)

WiFi Connection (Wireless ADB)

  1. Connect your phone via USB first
  2. Click the Device button in the top toolbar
  3. Click WiFi Connect
  4. Enter your phone's IP address (find it in Settings → WiFi → tap your network)
  5. Click Connect — you can now unplug the USB cable
WiFi ADB is slower than USB and may cause video lag. USB is recommended for production use, especially with phone farms.

Installing Portal APK

The Portal APK is a companion app that provides:

  1. Find the P button on your device panel
  2. Click the P button to install Portal
  3. Wait for installation (10-15 seconds) — button turns green when ready
  4. The K button should also turn green (keyboard active)
ButtonColorMeaning
PGreenPortal installed, accessibility ON, up to date
PYellowPortal installed but outdated — click to update
PRedInstallation failed — check USB debugging permissions
KGreenPortal IME keyboard active and ready
KRedIME not available — click to enable
If Portal installation fails: (1) Enable "Install via USB" in Developer Options, (2) For Samsung, enable "USB debugging (Security settings)", (3) Click the R button to reconnect.

4. Configuring LLM Providers & API Keys

DroidAI requires an LLM API key to power its AI agent. The AI reads the device screen and decides what actions to take.

Supported Providers

ProviderRecommended ModelsVisionNotes
AnthropicClaude Sonnet 4, Claude HaikuYesBest overall. Supports prompt caching for cost savings.
OpenAIGPT-4o, GPT-4o-miniYesGood alternative with fast response times.
Google GeminiGemini 2.0 FlashYesCost-effective option.
DeepSeekDeepSeek ChatNoBudget option, text-only (no screenshots).
GrokGrok-2YesxAI
Ollama CloudVariousVariesSelf-hosted models via Ollama.
KimiMoonshotNoChinese provider.
CustomAny OpenAI-compatible APIVariesUse with any OpenAI-format provider.

Getting an API Key

Anthropic (Recommended)

  1. Go to console.anthropic.com
  2. Create an account or sign in
  3. Navigate to API Keys in the dashboard
  4. Click Create Key and copy it (starts with sk-ant-)
  5. Add credits to your account (API is pay-per-use)

OpenAI

  1. Go to platform.openai.com
  2. Create an account or sign in
  3. Navigate to API Keys
  4. Click Create new secret key and copy it (starts with sk-)

Entering Your API Key in DroidAI

  1. Click the Settings button (gear icon) in the top toolbar
  2. In AI Model section, select your provider
  3. Paste your API key
  4. Select a model from the dropdown
  5. Click Save
Enable Prompt Caching (Anthropic only) to significantly reduce API costs — saves up to 90% on repeated requests.

Using a Custom Provider

If you have an OpenAI-compatible API endpoint (e.g., local model, proxy, third-party):

  1. Select Custom from the provider dropdown
  2. Enter the base URL (e.g., http://localhost:11434/v1 for Ollama)
  3. Enter your API key (if required)
  4. Type your model name manually

5. Interface Overview

DroidAI has a split-panel layout: Chat Panel on the left and Device Grid on the right.

Top Toolbar

ButtonFunction
DeviceDevice management: connect/disconnect, Portal, WiFi, restart ADB, resize
MirrorToggle mirror mode (green = ON). Forwards mouse/keyboard to device.
SettingsOpen settings (API keys, agent config, display, streaming)
ReportSubmit bug reports with diagnostic logs
User MenuSubscription status, email, sign out

Chat Panel (Left Side)

The chat panel has four tabs:

Activity

Live execution log. Shows commands, AI reasoning, tool calls, and results.

Presets

Manage saved commands, AI personas, AI rules, and app cards.

Advanced

Macros (record/replay), Workflows (multi-step chains), Triggers.

Log

Raw debug/diagnostic logs for troubleshooting.

Command Input Bar (Bottom)

ButtonFunction
StealthToggle stealth mode (red = ON). Human-like delays and jitter.
LoopToggle repeat/loop mode. Set count and interval.
ScreenshotCycle: OFF → AUTO → ALWAYS.
App CardsOpen app card grid for quick app-specific tasks.
All / NoneSelect or deselect all connected devices.

Device Grid (Right Side)

Each connected device appears as a panel with live screen mirror. Panel controls:

ButtonFunction
Selection badgeClick to select/deselect for commands
PPortal APK status & install
KKeyboard IME status & enable
RReconnect device
LOpen per-device activity log
EnlargeFull-screen device view
×Disconnect device

6. Your First Task — Quick Start

Let's walk through your very first DroidAI task, step by step.

Prerequisites Checklist

Example: Open YouTube and Search for a Video

  1. Make sure your device is selected (click its selection badge)
  2. Click the text input field at the bottom of the chat panel
  3. Type: Open YouTube and search for "lofi hip hop radio"
  4. Press Enter (or click the Send button)
  5. Watch the Activity tab — the AI will launch YouTube, tap search, type the query, and press Enter
You can watch the AI's actions in real-time on both the device screen and in the Activity tab.

More Example Commands

CommandWhat It Does
Open Instagram and like 3 posts in my feedLaunches Instagram, scrolls feed, likes 3 posts
Go to Settings and turn on WiFiOpens Settings, navigates to WiFi toggle
Open Chrome and go to google.comLaunches Chrome, navigates to URL
Send "Hello!" to Mom on WhatsAppOpens WhatsApp, finds contact, sends message

Stopping a Task

Click the red Stop button during execution, or press Ctrl+C while the input field is focused.

7. Core Features Deep Dive

How the AI Agent Works

For each step of a task, the AI agent follows this cycle:

  1. Observe — Read the UI tree and optionally take a screenshot
  2. Think — Analyze the screen state, decide the best action
  3. Act — Execute one action: tap, type, scroll, swipe, etc.
  4. Repeat — Loop until task complete or max steps reached

Available Actions

ActionDescriptionExample
tap(x, y)Tap at screen coordinatesTap a button
click(index)Click element by UI tree indexClick "Send" by index
type(index, text)Type text into input fieldType a search query
scroll(dir)Scroll up/down/left/rightScroll feed
swipe(dir)Swipe gestureSwipe through stories
long_tap(x, y)Long press at coordinatesOpen context menu
launch(pkg)Open app by package nameLaunch Instagram
back / home / enterSystem navigation buttonsGo back
wait(ms)Pause executionWait for content
doneMark task finishedTask complete

Planning Mode

When enabled, the AI generates a step-by-step plan before execution. Useful for complex tasks where you want to see the approach first.

Max Steps

Maximum actions per task (default: 30, configurable: 10-100). Prevents infinite loops. Increase for long tasks; decrease for simple actions.

8. Multi-Device & Phone Farm Setup

DroidAI is built for scale. Connect and control dozens of Android devices simultaneously from a single PC.

Hardware Setup

USB Hub Selection

For multiple phones, you need a powered USB hub:

HubPortsBest ForNotes
Sipolar A-42320Medium farm (10-20)Industrial grade, dedicated power supply
Sipolar A-40010Small farm (5-10)Compact, desk-friendly
Sipolar A-81230Large farm (20-30)Rack-mountable
Anker USB 3.07-13Small setups (3-7)Consumer-grade, reliable
Always use a powered USB hub with external power adapter. Unpowered hubs cannot supply enough current for multiple phones.

Recommended Phones for Farms

PhonePrice (Used)AndroidPros
Samsung Galaxy S8/S9$40-609-10Cheap, reliable ADB, common
Samsung Galaxy A series$50-8011-13Good value, newer Android
Google Pixel 3/4$50-7012-13Stock Android, fast ADB
Xiaomi Redmi Note$40-6011-13Budget, good specs
For best compatibility, use Samsung Galaxy S8/S9 or Google Pixel devices.

Connecting Multiple Devices

  1. Connect all phones to the USB hub via data cables
  2. Connect the USB hub to your PC
  3. Enable USB debugging on every device
  4. Authorize USB debugging on each phone
  5. Launch DroidAI — all devices should appear in the grid
  6. Install Portal on all: Device menu → Install Portal All

Sending Commands to Multiple Devices

Option 1: All Devices

Click "All" to select every device, then type your command. Each device gets its own independent AI agent.

Option 2: Selected Devices

Click the selection badge on each device to target. Commands run only on selected devices.

Performance Considerations

DevicesRecommended PCNotes
1-5Any modern PC, 8 GB RAMRuns smoothly
5-1516 GB RAM, decent GPULower resolution to 480p
15-3032 GB RAM, dedicated GPUUse 360p resolution
30+Multiple PCs recommendedSplit devices across PCs

9. App Cards & Playbooks

App Cards are pre-configured guides that help the AI understand specific apps. They contain navigation hints and app-specific tips that significantly improve accuracy.

How App Cards Work

When you send a command, DroidAI checks if the current app has an app card. If found, the card's instructions are injected into the AI's system prompt.

Using the App Card Grid

  1. Click App Cards button in the command bar
  2. A grid popup shows all available app cards (25+ pre-configured)
  3. Click any card to open the detail dialog
  4. View/edit card content, add instructions, then click Run

Pre-Configured Apps

Social Media

Instagram, Facebook, X (Twitter), Threads, TikTok, Snapchat, Reddit, Pinterest, LinkedIn

Messaging

WhatsApp, Telegram, Discord, KakaoTalk, LINE, WeChat

Media

YouTube, Spotify, Netflix

Utility

Chrome, Gmail, Google Maps, Play Store, Settings, Amazon, Naver

Creating Custom App Cards

  1. Go to Presets tab → App Cards
  2. Click New Card
  3. Enter the package name
  4. Write navigation tips and key UI elements
  5. Click Save

10. Mirror Mode

Mirror Mode lets you manually control a device using your PC's mouse and keyboard.

Enabling Mirror Mode

  1. Click Mirror in the top toolbar (turns green)
  2. Click on a device panel to select it
  3. Mouse clicks and keyboard input now go to the device

Controls in Mirror Mode

PC InputDevice Action
Left clickTap at position
Click and dragSwipe/drag gesture
Mouse wheelScroll on device
Keyboard typingText input
Right clickBack button

Mirror Mode + Multi-Device

When multiple devices are selected, input is forwarded to all selected devices simultaneously. Useful for setting up multiple phones with the same configuration.

Mirror Mode and AI agent commands can conflict. Turn off Mirror Mode before sending AI commands.

11. Loop & Repeat Mode

Loop Mode repeats the same command multiple times with optional delays between cycles. Essential for repetitive tasks.

Setting Up a Loop

  1. Click Loop button in the command bar
  2. Set Count: how many times to repeat (1-999)
  3. Set Interval: minutes between cycles (0-999)
  4. Type your command and send — it repeats automatically

Example: Social Media Engagement Loop

Loop: 10 cycles, 5 minute interval
Command: "Open Instagram, scroll feed, like 3 posts, then close the app"

Result: Every 5 minutes, each device opens Instagram,
likes 3 posts, and closes. Repeats 10 times over ~50 minutes.
Use longer intervals (5-15 min) for social media tasks to appear more natural.

12. Stealth Mode

Stealth Mode makes the AI's actions appear more human-like by introducing natural variations.

What Stealth Mode Does

FeatureDescription
Tap JitterRandom offset on tap coordinates (±12px)
Speed VariationRandom ±20% variation in action timing
Reading PausesRandom pauses (0.5-3s) between actions
Action DelaysVariable delays between consecutive actions

When to Use Stealth Mode

Stealth Mode is ON by default. For testing/debugging, turn it OFF for fast, precise actions.

13. Personas, AI Rules & Presets

Saved Commands

Save frequently-used commands for one-click execution:

  1. Go to Presets tab → Saved Commands
  2. Enter a name and command text
  3. Click Add
  4. Click Play next to any saved command to execute

AI Personas

Personas customize the AI agent's behavior. Only one persona can be active at a time.

Examples: Speed Runner (fast, skip verifications), Careful (verify before/after each action), Social Media Expert (navigate social apps expertly).

AI Rules

Rules are constraints always injected into the AI's system prompt. Multiple rules can be active simultaneously.

Examples: "Never purchase anything", "Close ads immediately", "Always use search instead of scrolling", "Skip sponsored content when liking".

14. Macros, Workflows & Triggers

Macros (Record & Replay)

Macros record AI actions and replay them without the LLM, saving API costs.

Recording a Macro

  1. Go to Advanced tab → Macros
  2. Click Record (turns red)
  3. Send a command and let the AI execute
  4. Click Stop Recording

Playing a Macro

Select the macro and click Play — replays exact actions without calling the LLM.

Macros replay exact coordinates. If the app UI has changed, the macro may fail. Re-record if needed.

Workflows

Chain multiple commands into sequential flows. Each step runs only after the previous one completes.

  1. Go to Advanced tab → Workflows
  2. Click New Workflow
  3. Add steps with commands
  4. Click Run to execute the chain

Triggers

Condition-based auto-execution. Set conditions and DroidAI evaluates them with the LLM to decide when to act.

  1. Go to Advanced tab → Triggers
  2. Create a trigger with condition and action
  3. Enable the trigger — DroidAI monitors and acts when conditions are met

15. Screenshot Modes & Vision

DroidAI can send screenshots to the AI for visual understanding. Three modes available:

ModeDescription
OFFUI tree only, no screenshots
AUTOUI tree + screenshot when tree empty or agent stuck (≥2 failures)
ALWAYSUI tree + screenshot every iteration

When to Use Each Mode

Screenshots increase API costs. Use AUTO for the best balance of accuracy and cost.
Non-vision LLM models automatically fall back to OFF mode.

16. Telegram Bot Remote Control

Control DroidAI remotely from your phone using a Telegram bot.

Setting Up the Telegram Bot

  1. Open Telegram and search for @BotFather
  2. Send /newbot and follow prompts to create a bot
  3. Copy the bot token
  4. Find your Chat ID via @userinfobot
  5. In DroidAI: Settings → Telegram Bot
  6. Paste token, enter Chat ID, enable Auto-start, click Save

Telegram Commands

CommandDescription
/helpList all available commands
/devicesShow all connected devices
/select [device]Set default device
/run [command]Execute a task
/screenshotTake and receive a screenshot
/repeat [n] [interval] [cmd]Loop execution remotely
/stopStop current execution
/statusCheck device and agent status

17. Settings Reference

AI Model

SettingDefaultDescription
ProviderAnthropicLLM provider selection
API KeyYour provider's API key
ModelSpecific model to use
Prompt CachingONCache system prompts (Anthropic only)

Agent

SettingDefaultRangeDescription
Max Steps3010-100Maximum actions per task
Action Delay0s0-5sPause between actions
Conversation History205-40Messages kept in context
Action History51-15Prior actions in prompt
Full Context31-5UI tree detail level
UI Tree FilterConciseConcise or Detailed output

Display

SettingDefaultDescription
LanguageEnglishUI language (13 languages)
Device Panel Size100%Scale device panels (40-200%)
Font Scale100%UI text size (50-200%)

Streaming (Scrcpy)

SettingDefaultRangeDescription
Resolution720p240-1080pDevice screen resolution
Bitrate4 Mbps1-12 MbpsStream quality
Max FPS305-60Frame rate

18. Troubleshooting & FAQ

Device Not Appearing

SymptomSolution
Device not listedCheck USB cable (data, not charge-only). Try adb devices.
unauthorized in adbCheck phone for USB debugging popup. Tap "Allow".
offline in adbUnplug and replug USB. Try different port.
Device appears then disappearsFaulty cable or USB port. Try different cable/port.

Portal Issues

SymptomSolution
P button stays redEnable "Install via USB" in Developer Options.
P button yellowClick P to reinstall latest version.
Accessibility not enablingManually: Settings → Accessibility → DroidAI Portal → ON.
K button redSettings → Language & Input → enable DroidAI Keyboard.

AI Agent Issues

SymptomSolution
Agent does nothingCheck API key. Verify internet. Check Activity for errors.
Agent taps wrong elementsEnable Screenshot mode (AUTO or ALWAYS).
Agent stuck in loopClick Stop. Try rephrasing your command.
"Max steps reached"Increase Max Steps in Settings (up to 100).
Text input failsEnsure Portal IME is active (green K).

Frequently Asked Questions

Q: How much does the LLM API cost?

A typical task (10-15 steps) costs ~$0.01-0.03 with Claude Sonnet, ~$0.005-0.01 with GPT-4o-mini. Screenshots add ~$0.01-0.02 each.

Q: Can I use DroidAI offline?

Internet is required for LLM APIs. However, you can use a local model via Ollama with the Custom provider for offline use.

Q: How many devices can I connect?

No hard limit. Practical limits depend on hardware. Users commonly run 10-30 devices per PC.

Q: Does it work with emulators?

Yes. Any ADB-compatible device works: BlueStacks, NoxPlayer, Android Studio AVD. Connect via adb connect localhost:PORT.

Q: Is my API key stored securely?

API keys are stored locally in settings.json in your AppData folder. Never sent to DroidAI servers.