Skip to content

How it works

ClawMobile Cloud runs a real Android runtime in the cloud and lets an agent drive it on your behalf. You never install anything — the phone, the agent, and the skill store all live server-side.

The big picture

Browser ──▶ Cloudflare Worker ──▶ Orchestrator ──▶ Android container
│ (auth, rate limit, SSE) (sessions, pool) (Redroid + plugin)
│ │
└────────── scrcpy stream ◀────────┘
  • Cloudflare edge — the Worker handles auth (magic-link + JWT), rate limits, and fans out session progress over SSE. The web UI is served from Pages.
  • Orchestrator — a Bun service that allocates a device, boots the OpenClaw gateway with the mobile-ui plugin, relays the screen, and syncs skills to R2 when the session ends.
  • Android container — a fresh Redroid instance per session. The agent controls it over ADB; you watch the screen through a low-latency scrcpy stream and can tap directly.

A session, step by step

  1. You request a session. The Worker checks your rate limit (1 active session, 5 per hour) and asks the orchestrator for a device.

  2. A device is allocated. The orchestrator starts a clean Redroid container on a Kamatera node — or bursts to a Vast.ai instance if the pool is full — and waits for ADB to come up. Progress streams to your browser live.

  3. The agent comes online. OpenClaw boots with the plugin in cloud mode. The phone screen appears; you can type instructions or tap the screen yourself.

  4. You demonstrate a task. Recording captures touches, screenshots, and app state. See Recording a demo.

  5. A skill is induced. On stop, the induction pipeline generalizes your demo into a reusable, versioned skill.

  6. The session ends. Skills sync to R2 (versioned), the container is destroyed, and its workspace is wiped. Nothing persists on the device.

What makes it safe to run unattended

Every container is single-use and isolated. A watchdog sweeps the pool every two minutes and destroys any container whose session has gone silent or any burst instance past its hard time cap — so a crash can never leave compute billing or your data sitting on a device.

Next: What are skills?