For developers

AI Agent Phone API

One small JSON API, over plain HTTP, to control a real Android device from your agent: read the screen, tap, type, swipe, launch, install. No per-app integration, no dedicated SDK required.

Early access · no payment, no card Plain HTTP, no lock-in

What the API gives you

Everything an agent needs to act like a user.

The same core actions whether your agent is testing an APK, filling out a form in an app with no website, or placing a call. Full reference →

Screen read
Get the current screen as a screenshot and as a structured accessibility tree, so the agent can reason about what's on screen, not guess at fixed coordinates.
Tap / long-press
Interact with any element by coordinate or by matched visible text.
Type text
Type into any focused field — forms, search boxes, chat composers.
Swipe / scroll
Navigate lists, feeds and multi-screen flows the same way a thumb does.
Launch / install
Start an installed app by package name, or install your own APK and apps from the real Google Play Store.
Screenshot evidence
Every action can be followed by a screenshot, so test runs and task completions are independently verifiable, not just self-reported.

Why a phone API

One API, instead of one integration per app.

Most mobile apps have no public API at all — the interface is the screen. A phone API sidesteps that entirely: instead of integrating with each app's backend, your agent drives the one surface every app already has — the touchscreen — through a single, stable API.

That means an agent built against the ERIUS PHONE API can work across any app installed on the device, from your own product under test to anything on the Google Play Store, without a new integration for each one.

Get started

Plain HTTP, API-key auth, any framework.

Give your agent a phone.

One API. A real Android device underneath it.

Request a device