Shipvo Simulators

110 phones on one machine.

Android and iOS simulators that run on your laptop or your CI runner. No cloud, no account, no meter. Your agent opens as many as the machine holds, and every one is its own phone.

$npx shipvo sim
Test in CI Read the benchmark
110simulators running

Each phone is a separate session with its own storage, network and device state. Measured on 2026-11-07: 110 open at once, every sampled one answering. The swarm above shows the 2-CPU, 7.8 GB run; a 4-CPU, 15.6 GB machine held 260.

110simulators open at once on a 2-CPU, 7.8 GB machine
57.1 MBof memory for each simulator
5.1 sor less to start each batch of 10, up to 100 open
260simulators open at once on a 4-CPU, 15.6 GB machine: write-up
10.7 / 10.2 sto the first tap on Android / iOS, from a clean machine
How it works

One browser. One phone per session.

Shipvo builds a web copy of your app and serves it to a headless Chromium. Each session gets its own browser context in an Android or iOS phone frame, so a hundred phones cost a hundred small contexts, not a hundred virtual machines.

agent ortest runner shipvo simlocalhost one Chromium
  1. 01
    Isolated by default

    Every session has its own storage, cookies, network profile and device state. Sessions share nothing.

  2. 02
    The app you ship

    The simulator runs the web build of your app, framed as an Android or an iOS phone with its own screen size and system sheets.

  3. 03
    One API for agents and tests

    Tap, type, scroll, read the screen, set the device, push a notification. Over MCP for agents, or from typed tests in CI.

Measured

Drag it. Every number is a run.

We opened sessions in batches of 10 and read the machine after each batch. Move the slider to see what the machine gave up at every step.

Open
110
Memory used
0
Batch start
0

Memory used (GB) against simulators open. Raw numbers.

Local and CI

The same simulator on your laptop and in your pipeline.

Start it with one command. Point your agent or your tests at it. On a clean GitHub-hosted runner with only Node, the first tap lands about ten seconds after the process starts, on Android and on iOS.


      

Android

start 0.00server 8.99app loaded 10.42first tap 10.73 s

iOS

start 0.00server 8.81app loaded 9.88first tap 10.19 s

Measured on 2026-11-07 on a GitHub-hosted ubuntu-latest runner. CI run 42355471525, script scripts/first-run/run-time.mjs.

Android and iOS

Both platforms. Same app. Side by side.

Choose shipvo run android or shipvo run ios. Each phone has its own size, screen cut-out and system sheet, and your tests run on both.

Android

A punch-hole phone with a gesture bar.

Size
412 x 915
Look
Punch-hole camera, gesture bar
System sheet
Android system sheet

iOS

An iPhone with a Dynamic Island and a home indicator.

Size
393 x 852
Look
Dynamic Island, home indicator
System sheet
iOS system sheet

100% actionable parity on all 13 screens, Android and iOS

For agents

Your agent spawns its own phones.

Add the simulator to Claude Code, Cursor, Codex or any agent that speaks stdio MCP with one command: npx shipvo agents setup. The agent calls open_app to get a phone, list_sessions to see them all and end_session when it is done.

Free

Apache-2.0 open source. No per-simulator fee, no seat, no minutes to buy.

On your machine

The simulator runs on your computer. Your app, your data and your test runs stay there.

In your CI

Runs on a standard GitHub-hosted runner, for Android and for iOS.

Open your first hundred.

Make an app, start the simulator, hand it to your agent.

$npm create shipvo@latest
Get started Agent docs All features