Name: opendroid
Author: yashab-cyber

安装

添加到你的 Claude Code skills 目录：

# Add to your Claude Code skills
git clone https://github.com/yashab-cyber/opendroid

安全报告

已验证

上次扫描：—

{
  "status": "PASSED",
  "issues": []
}

README.md

🎯 What is OpenDroid?

OpenDroid isn't just another chatbot. It's a fully autonomous AI agent that lives on your Android phone and actually does things for you.

"Check if it's going to rain tomorrow, and if so, text my wife that I'll be late and set an alarm for 6 PM."

OpenDroid will plan this as 3 steps, execute each one, verify the results, and adapt if anything fails — all without you lifting a finger.

✨ Features

🧠 Autonomous Agent Engine

Capability Description

Self-Planning Breaks complex commands into sequential steps with dependency tracking

Habit & Routine Detection Mines recurring daily app sequences (e.g. Gmail $\to$ Calendar $\to$ Slack at 9 AM) and offers one-click automated routines

Re-Evaluation Monitors execution results and dynamically replans when steps fail

Compound Intent Guard Smart detection of multi-action commands (e.g. "open WhatsApp and send message")

Contact Disambiguation 4-tier contact resolution with fuzzy matching and relationship aliases ("call dad")

🛠️ On-Device Model Manager (LiteRT-LM)

Capability Description

Background Downloader Real network downloads (via WorkManager) with Pause/Resume, cellular support, speed tracking, and ETA

Secure Authentication Direct Android Keystore AES-GCM token storage to securely fetch gated Hugging Face models

Integrity Verification Computes SHA-256 hashes and verifies LiteRT engine loading compatibility before marking READY

Local Model Import Direct offline importing of catalog or freestanding custom .task / .litertlm files with JNI verification checks

📱 Full Device Control

Action Examples

System Brightness, WiFi, Bluetooth, Flashlight, DND, Volume, Screenshot

Communication WhatsApp messages, Telegram messages/channels (@user), Calls, SMS, Email drafts

Routines Morning briefings, agenda summaries, automated macro sequences

Productivity Screen extraction ("Read & Remember"), Alarms, Timers, Reminders, Calendar events, Notes

Navigation Google Maps directions, Uber/Ola booking

Media Play/pause music, YouTube search, camera

Finance UPI payments, bill splitting, currency conversion

Smart Home Google Home device control

👁️ Vision Engine & Screen Understanding

Captures screenshots via Accessibility API and feeds them to vision-capable LLMs for real-time screen analysis and "Read & Remember" multimodal meeting/note extraction. Falls back to accessibility tree text-scraping on older devices.

🗄️ 4-Tier Personal Knowledge Graph

┌─────────────────────────────────────────────────────────────┐
│                 Personal Knowledge Graph                    │
├──────────────┬──────────────┬────────────────┬──────────────┤
│ ⚡ Level 1   │  🧠 Level 2  │   📈 Level 3   │  🔒 Level 4  │
│  Temporary   │  Long-Term   │    Learned     │  Sensitive   │
│   (Active    │  (Explicit   │    Patterns    │ (Keystore    │
│    Plan)     │    Facts)    │  (Inferences)  │  Encrypted)  │
└──────────────┴──────────────┴────────────────┴──────────────┘

🎙️ Voice Interface

Offline wake word detection — say "OpenDroid" to activate
Speech-to-text for hands-free commands
Text-to-speech with ElevenLabs premium voice support

🎨 Premium UI

Built with Jetpack Compose featuring a futuristic glassmorphic design:

Deep navy (#080C10) + Neon green (#00FF88) color system
Pulsing audio orb animation during listening
Live latency benchmarks for each provider
Dark mode by default

🎬 Meet OpenDroid in 3D

Note: If the video doesn't play inline on GitHub, click here to download and watch it.

📸 Screenshots

🏗️ Architecture

Clean architecture with Dagger-Hilt dependency injection:

com.opendroid.ai
│
├── 🤖 accessibility/      App automators (WhatsApp, SMS, Calls)
├── ⚡ actions/             60+ action executors across 10 modules
├── 🧠 core/
│   ├── agent/              AgentLoop, PlanManager, IntentClassifier, VisionEngine
│   ├── llm/                12 LLM providers, fallback chain, prompt engine
│   ├── memory/             4-tier memory system + notification intelligence
│   ├── security/           Direct Keystore provider credentials + legacy plaintext migration
│   ├── service/            Foreground service, notification listener, boot receiver
│   └── voice/              Wake word, speech recognition, TTS engine
│
├── 💾 data/
│   ├── db/                 Room database (7 DAOs, 7 entities, 3 migrations)
│   ├── models/             Unified data models (Plan, Memory, ChatMessage)
│   └── repository/         Repositories backed by Room & DataStore
│
├── 💉 di/                  Hilt modules (App, Database, LLM)
└── 🎨 ui/
    ├── theme/              Glassmorphic design system
    ├── screens/            16 screens (Chat, Plan, Memory, Settings, etc.)
    ├── viewmodel/          8 ViewModels
    └── components/         Reusable Compose components

🔌 Supported LLM Providers

OpenDroid supports 12 LLM providers with automatic failover:

Provider Models Type

🟢 Google Gemini Gemini 2.0 Flash, Pro, Nano Cloud + On-device

🟣 Anthropic Claude Claude Sonnet 4, Opus 4 Cloud

🔵 OpenAI GPT-4o, GPT-4.1, o3 Cloud

⚡ Groq LLaMA 3, Mixtral (ultra-fast) Cloud

🔷 DeepSeek DeepSeek V3, R1 Cloud

🟠 Mistral AI Mistral Large, Medium Cloud

🌐 OpenRouter 200+ models via unified API Cloud

🤝 Together AI Open-source model hosting Cloud

🔴 Cohere Command R+ Cloud

🐙 GitHub Copilot GPT-4.1, Claude via Copilot API Cloud

🏠 Ollama Any local model (LLaMA, Phi, etc.) Local

🔧 Custom OpenAI Any OpenAI-compatible endpoint Self-hosted

Smart Fallback: If your primary provider fails, OpenDroid automatically tries the next available provider in the chain.

⚡️ Getting Started

Prerequisites

JDK 21 — not newer. The project compiles against Java 21 (jvmToolchain(21)). gradle/gradle-daemon-jvm.properties makes ./gradlew select an installed JDK 21 automatically, so you do not have to change JAVA_HOME — but a JDK 21 must be installed.
Android SDK 35 (Android 15)

Build & Install

# Clone the repository
git clone https://github.com/yashab-cyber/opendroid.git
cd opendroid

# Build debug APK
./gradlew assembleDebug

# APK output location
# → app/build/outputs/apk/debug/app-debug.apk

Required Permissions

On first launch, OpenDroid will guide you through granting:

Permission Why

🔓 Accessibility Service UI automation, screen reading, app control

⚙️ Write Settings Toggle WiFi, Bluetooth, brightness

🎤 Record Audio Wake word detection & voice commands

🔔 Notification Access Smart notification reading & auto-reply

📱 Post Notifications Foregr

常见问题

What is opendroid?⌄

opendroid is an open-source ai agents skill for AI coding assistants such as Claude Code, Codex CLI, and ChatGPT, built by yashab-cyber. Your Open Autonomous Android Agent — A production-ready, self-planning AI assistant powered by local/remote LLMs and accessibility-driven screen automation. It has 907 GitHub stars.

Is opendroid safe to use?⌄

Yes. opendroid passed SkillsLLM's automated security scan — a dependency vulnerability audit plus prompt-injection heuristics — with no high-severity issues. You can read the full report in the Security Report section on this page.

How do I install opendroid?⌄

Clone the repository with "git clone https://github.com/yashab-cyber/opendroid" and add it to your Claude Code skills directory (see the Installation section above).

What programming language is opendroid written in?⌄

opendroid is primarily written in Kotlin. It is open-source under yashab-cyber on GitHub, so you can review or fork the full source.

Are there alternatives to opendroid?⌄

Yes. SkillsLLM lists many other AI Agents skills you can browse and compare side by side. Open the AI Agents category from the badge at the top of this page, or use the Related Skills and comparison links further down to weigh opendroid against similar tools.

opendroid

安装

快速入门