Guide users on their screen, right inside your app.
Live in-app screen streaming to your agent console, pointer guidance, and permissioned semantic actions. The device validates every command; raw touch injection doesn't exist here.
Three steps from “I'm stuck” to solved
Start sharing
User taps “Get live help”
A consent sheet explains exactly what the agent will see. Capture starts only after the user agrees, never before.
Agent sees and points
Live video with a laser pointer. Sensitive areas are masked on-device, so redacted pixels never reach the stream.
Act, with permission
A second consent unlocks control. Agents invoke only the semantic actions you registered; the device validates and acknowledges each one.
The device is authoritative. Always.
Remote support tools usually work by injecting touches. ScreenSteer refuses that entire class of risk. There is no code path for it.
Everything a support session needs
Try the two-window demo
Open the device simulator in one window and the agent console in another. Grant consent, point, request control, invoke an action, and watch the acknowledgement come back with measured latency.
+ consent sheet
actions + timeline
Flat plans. Unlimited sessions and apps.
Register actions. Ship the SDK. Done.
A typed wire protocol with a device-side validator, reliable and lossy data channels, and layout deltas. Choose Web, Windows, Android, Flutter, iOS, or React Native, with maturity and requirements stated in each guide.
ScreenSteer.register("workorders.refresh") {
// your code: the only thing an agent can run
await store.refreshWorkOrders()
}
ScreenSteer.redact(accountPanel)