callstackincubator

CLI to control iOS and Android devices for AI agents

896
55
100% credibility
Found Feb 04, 2026 at 301 stars 3x -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
TypeScript
AI Summary

Command-line tool for capturing accessibility snapshots and automating taps, typing, and gestures on iOS simulators and Android emulators/devices to enable AI agent control.

How It Works

1
🔍 Discover agent-device

You hear about a handy tool that lets AI helpers control phone screens like a real person.

2
📥 Get it set up

With one easy command, you add the tool to your computer and it's ready to go.

3
📱 Pick your phone

Choose the phone simulator or real device you want your AI to play with.

4
📸 Snap a picture of the screen

The tool grabs a clear map of everything on the screen, labeling buttons and fields perfectly.

5
🖱️ Tell it what to do

Point to a button to tap, type in a box, or scroll around just by describing it.

6
Watch the magic happen

Your AI sees the screen and acts just like you would, smooth and reliable.

AI takes full control

Now your smart assistant handles phone tasks on its own, saving you time and effort.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 301 to 896 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is agent-device?

Agent-device is a TypeScript CLI for letting AI agents control iOS simulators/devices and Android emulators/devices, pulling from Vercel's agent-browser ideas. Developers get commands like `open SampleApp`, `snapshot` for accessibility trees, `click @e5`, `fill @e6 "John"`, or `scrollintoview "Wi-Fi"` to automate UI flows without heavy frameworks. Run via `npx agent-device` on Node 22+, handling sessions, traces, and replays for agent-driven exploration.

Why is it gaining traction?

It stands out with zero-build TypeScript execution, minimal deps, and npx-ready CLI control for quick agent device testing—no Xcode/Appium installs needed for basics. Features like semantic `find label "Wi-Fi" click`, ref-based interactions (`@e5`), and replay updates fix selector drift automatically. 502 stars reflect appeal for AI agent device workflows, plus GitHub Actions/Copilot integration via CLI github commands.

Who should use this?

AI engineers building mobile agents for app testing or RPA, mobile devs scripting UI flows without Detox/EarlGrey overhead, or QA teams needing fast simulator automation. Ideal for angular cli control flow migration or amp device agent high cpu debugging via snapshots.

Verdict

Grab it for experimental AI agent device control—CLI github ubuntu/windows/linux support shines—but 1.0% credibility score flags early-stage risks; solid README offsets thin tests. Stars at 502 suit prototypes, not prod; PRs welcome to mature it.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.