DroidClaw
Automates Android phones to complete plain-English goals by reading the screen, reasoning with an LLM, and tapping and typing via ADB until done
The product is an AI agent that automates Android phones by reading the screen via the accessibility tree, reasoning with an LLM, and executing taps, types, and swipes via ADB in a loop until a plain-English goal is complete. It is for developers and operations teams in B2B contexts who need to automate mobile apps without APIs. Unlike predefined button flows, it adapts to UI changes and popups, and offers interactive goals, AI-powered JSON workflows across apps, and deterministic YAML flows without LLM calls.
Key features
- Reads screen via UIAutomator accessibility tree
- Parses tappable elements with coordinates
- Executes actions via ADB
- 22 actions: tap, type, swipe, launch
- Stuck recovery after 3 unchanged steps
- Screenshot fallback for empty tree
- Interactive natural-language goals
- Workflows chain goals across apps (JSON)
- Flows deterministic YAML without LLM
- Remote control via Tailscale ADB
- Cron job workflow scheduling
- Send WhatsApp to saved/unsaved numbers
- Reply to latest SMS
- Compose emails via Gmail
- Post messages to Slack and Telegram
- Search Google and collect results
- Ask ChatGPT/Gemini and grab answers
- Check weather, stocks, flights
- Compare prices across apps
- Translate via Google Translate
Social posts
- No social media activity within the last 30 days
GTM channels
- Community
ICP
- Software developers
- DevOps sre teams
- Engineering teams