← Back to research
·6 min read·company

droidclaw

droidclaw turns old Android phones into AI agents via ADB — give it a goal in plain English, it reads the screen, reasons about what to do, taps and types, and repeats until the job is done. No APIs needed.

Key takeaways

  • Turns any old Android phone into an autonomous AI agent via ADB screen reading and interaction — no app APIs needed
  • Perception → reasoning → action loop with stuck detection, repetition tracking, drift detection, and vision fallback
  • Can delegate tasks to ChatGPT, Gemini, or Google Search apps on the device without API keys — uses apps like a human would
  • TypeScript/Bun runtime with accessibility tree parsing and optional screenshot-based vision for webviews and Flutter apps

FAQ

What is droidclaw?

droidclaw is an AI agent that controls Android phones via ADB. You give it a goal in plain English, and it reads the screen, decides what to tap or type, executes via ADB, and repeats until the goal is complete.

How does droidclaw work?

A perception → reasoning → action loop: dump the accessibility tree via ADB, send screen state + goal to an LLM, execute the returned action (tap, type, swipe), and loop until done.

How much does droidclaw cost?

Free and open-source. You need an old Android phone, a computer running Bun, and an LLM API key.

Who competes with droidclaw?

No direct competitors in the personal agent space — most agents use APIs, not screen reading. Conceptually similar to browser automation agents like HappyCapy but for Android devices.

Executive Summary

droidclaw turns old Android phones into autonomous AI agents. Instead of building API integrations, it controls phones the way a human would — reading the screen via ADB accessibility trees, reasoning about what to do with an LLM, and executing taps, types, and swipes. No app APIs, no custom integrations — just install apps and tell the agent what you want done. [1]

Status (as of June 2026): Stalled. The repo is not archived and the website and hosted dashboard remain live, but there have been no commits since February 28, 2026 and no releases since v0.5.3 (February 25, 2026) — roughly three and a half months of silence in a fast-churn ecosystem. [2]

AttributeValue
Authorunitedbyai
LanguageTypeScript (Bun)
LicenseOpen source (MIT per website; no LICENSE file detected in the repo)
GitHub Stars~1,523 (as of June 2026)
Latest Releasev0.5.3 (February 25, 2026)
FundingNot publicly disclosed

Product Overview

droidclaw's core innovation is using the phone itself as the integration layer. Rather than building API connectors for every service, it reads screens and interacts with any installed app. It can even delegate questions to ChatGPT, Gemini, or Google Search on the device — no API keys needed for those services. [1]

How It Works

  1. Perceive — Dump accessibility tree via ADB, parse interactive UI elements, diff with previous screen
  2. Reason — Send screen state + goal + history to LLM, get back think/plan/action
  3. Act — Execute via ADB (tap, type, swipe), feed result back on next step
  4. Loop — Repeat until goal is done or step limit reached

Reliability Features

FeatureDescription
Stuck Loop DetectionRecovery hints after 3 unchanged screens
Repetition TrackingSliding window catches retry loops across screen changes
Drift DetectionNudges agent if it spams navigation without interacting
Vision FallbackScreenshots for webviews, Flutter apps, games (empty accessibility trees)
Action FeedbackEvery action result fed back to LLM for next step
Multi-Turn MemoryConversation history maintained across steps

Execution Modes

Beyond interactive goals, droidclaw added two structured execution modes: Workflows chain multi-app tasks defined in JSON with AI reasoning at each step, and Flows run deterministic YAML scripts instantly with no LLM calls at all. It supports ADB over WiFi and multiple LLM providers (Groq, Ollama, OpenAI, Bedrock), and the v0.5.x line shipped a companion APK and a hosted web dashboard at app.droidclaw.ai. [3]


Strengths

  • Universal integration — Works with any Android app without APIs
  • Hardware recycling — Gives old phones a second life as AI agents
  • Robust failure handling — Stuck detection, repetition tracking, drift detection, vision fallback
  • No API keys for target apps — Uses ChatGPT, Gemini, Google Search as a human would
  • Web dashboard — Visual monitoring at app.droidclaw.ai, still live as of June 2026
  • Real traction — ~1,523 stars and 225 forks as of June 2026, with a companion APK for device-side setup

Cautions

  • Android only — No iOS support (ADB is Android-specific)
  • Inherently fragile — Screen-based automation is slower and less reliable than API calls
  • Requires ADB setup — USB debugging, computer running Bun
  • LLM latency — Each step requires an LLM call; multi-step tasks are slow
  • Bun-only — Won't run on Node.js; uses Bun-specific APIs
  • Stalled maintenance — No commits since February 28, 2026 and no release past v0.5.3 (February 25, 2026) as of June 2026; in the fast-churn OpenClaw ecosystem, that pause is a real abandonment risk [2]
  • License ambiguity — The website calls it MIT, but the repo has no detectable LICENSE file [1]

What Developers Say

Despite strong GitHub traction, substantive first-hand community commentary is scarce. The February 2026 Show HN launch ("Show HN: DroidClaw – Turn old Android phones into AI agents") drew only 3 points and zero comments [4], and we found no verbatim developer reviews on Reddit or X as of June 2026 — most coverage is launch-announcement amplification and tutorial walkthroughs rather than usage reports. [5]


Pricing & Licensing

TierPriceIncludes
Open SourceFreeFull agent; MIT claimed on the website, though the repo lacks a LICENSE file

Hardware cost: Any old Android phone + computer. Hidden costs: LLM API usage per step (Groq has a free tier; other providers charge per token). [3]

Funding: Not publicly disclosed — no announced investors or commercial entity behind the unitedbyai org.


Competitive Positioning

CompetitorDifferentiation
OpenClawOpenClaw uses API integrations; droidclaw uses screen reading — no APIs needed
HappyCapyHappyCapy automates browsers; droidclaw automates Android phones
ManusManus is cloud-managed; droidclaw is self-hosted phone control

Bottom Line

droidclaw is a genuinely novel approach to personal AI agents: instead of building integrations, use the phone's screen as the universal API. It's slower and more fragile than API-based agents, but it works with any app without any setup. The "turn old phones into agents" pitch is compelling for hardware recycling and for automating apps that have no API. The caveat as of June 2026: development stopped in late February after a two-week burst, so treat it as a promising snapshot rather than a maintained project until commits resume.

Recommended for: Tinkerers with spare Android phones who want to automate apps without APIs and can tolerate an unmaintained codebase.

Not recommended for: Production workflows, time-sensitive automation, or users wanting reliable API-based integrations — or anyone who needs an actively maintained project.


Research by Ry Walker Research • methodology