ai agent for real phone touch

[Mark Harrison]

Hi there,

I found your GitHub through the agent automation side around OpenClaw, and it seemed close enough to what I’m working on to reach out.

I’m Mark Harrison, building Lumi, a small early hardware project for AI agents.

Agents are getting better in browsers, terminals, and desktops, but a lot of real workflows still hit a wall inside mobile apps. Lumi is a local-first dock that lets an agent observe and operate a real phone with taps, swipes, and typing, without depending on app APIs or custom integrations.

No pressure at all , I’m mainly looking for honest early feedback from people building with agents. Would it be okay if I send you a short preview?

Best,
Mark Harrison

[Ethan Reed]

Hi Weitai,

I came across your GitHub from the OpenClaw / agent automation ecosystem, and it seemed like the kind of context where a phone-use agent project might be worth sharing.

I’m Ethan Reed, and I’m building Lumi, a local-first hardware project for AI agents.

It feels like agents are getting much better in browsers, terminals, desktops, and APIs, but mobile apps are still a blind spot when important workflows live on a real phone. Lumi is a small hardware direction that lets agents observe and operate a real smartphone through physical taps, swipes, and typing.

I’m putting together a short technical preview and would really value your feedback. Would it be okay if I send it over when it’s ready?

Best,
Ethan Reed