𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎

@agent_device

The mobile verification for AI Agents. By @callstackio

Poland

agent-device.dev

Replies Reposts Media

𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎
@agent_device
Sep 1
Connect to Expo Simulators with agent-device ⬇️
Khalid
@KhalidDevLog
Sep 1
Just tried @expo Simulators in the browser and it actually changes how you develop apps. No Xcode, no setup, no waiting. Run your app on a real iOS simulator from anywhere. Thank you @expo for building this, and thank you @joe_d_ryan for giving me the chance to try it. 🙏
4
𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎
@agent_device
Aug 26
Single prompt, everyday, let agent find bugs for you: » use agent-device help dogfood <this app>
Michał Pierzchała
@thymikee
Aug 26
I get a better experience asking Claude to find the bugs for me. Single prompt was enough for Opus 5 to freely explore the Wikipedia Android app, catch a visual glitch, and flag it itself. You can do this today with any agent.
00:00
𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎
@agent_device
Aug 18
token efficiency = cost efficiency. why pay more when you can pay less
Michał Pierzchała
@thymikee
Aug 18
agent-device delivers more finished work for every dollar. Benchmarked on @swmansion’s AppControlBench with GPT-5.4-mini and Haiku 4.5: → ~4× more completed tasks/$ → 94.2–98.3% task completion → 2% faster with GPT-5.4-mini; 26% slower with Haiku 4.5 than alternative
𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎
@agent_device
Aug 7
4k stars on GitHub 572k monthly downloads on npm
𝚊𝚐𝚎𝚗𝚝-𝚍𝚎𝚟𝚒𝚌𝚎
@agent_device
Aug 6
We didn't like that at the beginning, but data convinced us this is a good default. Not best at every case, that's why it's a flag that agents can chose to run
Michał Pierzchała
@thymikee
Aug 6
We cut the agent loop in half without making it slower. Old: snapshot → act → snapshot → act → … New: snapshot → act --settle → act --settle → … ⊙ 30% fewer steps ⊙ 50% less variance ⊙ Same total time ⊙ Fewer tokens @agent_device just got a lot more reliable
00:00