Index / DroidClaw · updated Oct 24, 2026
DroidClaw
unitedbyai/droidclaw · healthy · rank 28 of 50 by stars
Turns old Android phones into autonomous AI agents that read the screen via accessibility trees and tap/type through ADB. Its killer trick is delegating tasks to on-device apps like ChatGPT or Gemini without needing API keys.
Facts
Repository
Runtime
Posture
Community
Security breakdown
Composite 45 / 100 · how these are scored
higher is safer
higher is safer
higher is safer
higher is safer
higher is riskier
Model access
Read from the repository, not written by a model · 29 files examined
3 providers · runs locally · gateway support
Pinned models
llama-3.3-70b-versatile released 2024-12-06 — 23 months old (from the public model catalogue)
Pin last edited 8 months ago
Several defaults are pinned; the project picks one per provider rather than shipping a single default.
Evidence
Decision
Why choose DroidClaw over OpenClaw?
Why choose this
- Controls real Android UIs directly, no app integrations needed
- Can use on-device ChatGPT/Gemini apps without API keys
- Gives old phones a second life as cheap agent hardware
Tradeoffs
- No multi-channel messaging surface like OpenClaw's 20+ integrations
- Early-stage (v0.5.x) with active bugfix churn and no stated license
- ADB-level device control carries higher operational and security risk
Best fit
- Repurposing spare Android phones as AI agents
- Automating mobile apps that have no API
- Experimenting with perception-reasoning-action agent loops
Avoid if
- You need strong sandboxing or enterprise security controls
- You want a mature, stable production release
- You don't have an Android device or ADB setup available
Evidence is moderate: the README and recent commits give a clear picture of architecture and activity, but there are zero Reddit mentions and web sentiment is mostly about OpenClaw generally, not DroidClaw specifically. License and security posture are uncertain.
AI layer reviewed Oct 5, 2026 · how this is written
Star activity
1,594 stars today
Overview
DroidClaw is an open-source AI agent that turns Android phones — including old, spare devices — into autonomous operators. You give it a goal in plain English, and it runs a perception → reasoning → action loop: it dumps the accessibility tree via ADB, parses the XML into interactive UI elements, diffs against the previous screen to detect changes, optionally captures a screenshot, then asks an LLM what to do next. Actions include tapping, typing, swiping, and launching apps, repeated until the goal completes or the step budget runs out.
Its standout capability is delegating work to on-device apps like ChatGPT, Gemini, or Google Search and bringing results back — no API keys required, since it drives the apps the way a human would. The project ships an Android APK, a web dashboard, scheduled goals (via QStash), and a Discord community. Recent commits show active development: a comprehensive architecture guide covering all 26 actions, pairing flow, and failure recovery, plus fixes to session management, stop-button handling, and database migrations.
Compared to OpenClaw, DroidClaw is narrower but more tactile: instead of meeting you in chat channels, it physically operates a phone's UI, which makes it powerful for apps without APIs but riskier from a security standpoint (full device control via ADB, no sandboxing story, and no declared license). It's best suited for tinkerers who want to experiment with mobile UI agents rather than users seeking a polished, multi-channel personal assistant.