Pinky: Open-Source Tool Lets AI Agents Control Phones Wirelessly
Pinky is an open-source project that enables AI agents to control iOS and Android phones via Wi-Fi, supporting cross-platform desktop integration, secure approval workflows, background operation on Android 13+, and detailed operation logging, demonstrated by hands-on tests posting videos and earning rewards.
Project Overview
Developer DIYgod announced Pinky , an open-source project that lets any AI agent control iPhones and Android phones via Wi‑Fi, with full support for macOS, Windows, and Linux. The announcement on X garnered 220,000 views and the GitHub repository surpassed 1,000 stars within 25 hours, followed by 30 releases in a single day as the author responded to community feedback.
Key Differentiator: Wireless Connection
Unlike earlier tools such as iPhone Use that require a permanent USB cable between the phone and the computer, Pinky supports Wi‑Fi pairing. After an initial USB connection to authorize USB debugging (and wireless debugging on Android 11+), the phone can be controlled wirelessly as long as both devices are on a reachable network.
Setup Process
Install the desktop client – download the appropriate package from the GitHub Releases page for your OS.
Connect the phone – on Android, enable Developer Options and USB debugging, connect via USB once, approve the debugging prompt, then optionally switch to wireless debugging (Android 11+). Pinky then prompts installation of the Pinky Companion helper app to accelerate UI reading and text input.
Integrate with an AI agent – Pinky exposes phone capabilities through the Model Context Protocol (MCP). Clicking “Connect” next to a supported agent (Claude Code, Codex, Cursor, or any MCP‑compatible agent) automatically configures the integration.
Platform‑Specific Behavior
Android 13+ : Pinky uses Android’s background display feature to create a separate virtual screen. The agent can open apps, read UI, and tap buttons while the physical screen stays off and locked, allowing the user to continue using the phone normally.
iOS : No secondary screen exists, so the agent operates on the iPhone’s main display.
Hands‑On Test 1: Post a Kuaishou Video
“Use Pinky to post a video on Kuaishou Speed Edition with the first photo from my album and the caption: ‘This cup is really useful.’”
The agent drove the phone step by step, with each action mirrored in a desktop window. The entire flow completed in approximately 4 minutes and the video was published successfully.
Hands‑On Test 2: Watch Videos to Earn Coins
“Help me use Pinky on Kuaishou Speed Edition to watch videos and earn coins once.”
Starting from a wallet balance of 0 coins, the agent opened the app, navigated to the video feed, watched videos, followed on‑screen coin‑task prompts, and checked the rewards page. This multi‑step, conditional workflow took about 7 minutes and yielded 1,222 coins , demonstrating the agent’s ability to chain decisions across screens rather than executing fixed coordinates.
Security & Safety Features
Approval layer independent of the AI agent : Payments, transfers, and transactions require manual confirmation by default.
Configurable approvals : Users can also require approval for sending messages, deleting content, making calls, installing apps, etc.
Sensitive app isolation : Banking, wallet, and password‑manager apps are hidden from the agent; password and OTP fields cannot be filled by the agent.
Full audit trail : Every agent step is recorded as text, screenshots, and screen video, exportable as an HTML report for later review.
Manual takeover : The user can click the phone mirror at any time to seize control, pausing the agent.
Developer Use Case: Automated Mobile App Testing
Pinky enables coding agents to install self‑built APKs, launch them, collect runtime logs and crash reports, diagnose issues, modify code, and reinstall for iterative testing — all without human intervention.
Remote Access via Tailscale
Pinky supports Tailscale for connecting to Android phones over the internet, removing the requirement that the phone and computer share a local network.
Caution
Some apps detect accessibility services and automation; the author advises respecting platform rules and avoiding bulk reward farming or restriction bypasses.
Signed-in readers can open the original source through BestHub's protected redirect.
This article has been distilled and summarized from source material, then republished for learning and reference. If you believe it infringes your rights, please contactand we will review it promptly.
SmartApe AI
SmartApe's AI technology content account, focusing on cutting-edge AI developments and open-source frontlines.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.
