The virtual phone platformbuilt for AI automation

VMOS Edge runs Android virtual phones on local devices and combines AI Agents, workflows, and API control to help teams handle app testing, account management, data processing, and mobile automation more efficiently.

VMOS Edge Desktop
VMOS Edge Desktop Screenshot
Core Advantages

What is VMOS Edge?

VMOS Edge is a virtual-phone platform that runs on local devices, creating multiple independent Android environments at once. With AI Agents, automation scripts, and API control, it provides a stable, secure, and controllable runtime for app testing, social-media operations, account management, and more.

AI operates the phone

AI reads the phone screen and performs taps, input, and app launches automatically.

Many phones in parallel

Run multiple independent Android environments locally — ideal for batch testing and multitasking.

Runs locally, more control

Tasks and data run on local devices, reducing cloud dependence for higher stability.

Automation-ready

Turn repetitive actions into reusable flows to cut manual work and boost efficiency.

Company Business

Android cloud phones and AI automation for local deployment

VMOS Edge provides locally deployed Android virtual-phone platforms, edge-computing hardware, desktop management clients, automation APIs, and AI Agent integrations for developers, testing teams, and business operations. We help customers run app testing, batch device management, mobile workflow automation, isolated account environments, teaching demos, and data processing in controllable local environments.

Virtual-phone platform

Run multiple independent Android environments in parallel with multi-version systems, file management, screen control, and environment isolation.

AI automation

Connect taps, input, screenshots, verification, and other mobile actions to AI toolchains through APIs, SDKs, MCP, and workflow tools.

Local hardware deployment

VMOS L1, K1, and ARM server options cover local compute needs from small teams to scaled operations.

AI Automation

Let AI Agents operate the virtual phone directly

Let AI Agents operate the virtual phone directly

No need to hard-code scripts in advance — just describe the task goal, and the AI recognizes the screen, plans the steps, performs the actions, and returns the results.

Natural-language tasking

Just describe the goal; the AI breaks it into steps and executes.

Observe while acting

Reads page state in real time and adjusts the next actions.

Adapts to UI changes

Keeps completing the task through pop-ups, redirects, or branches.

Results returned automatically

Returns screenshots, a summary, and key results when done.

Generate automation scripts in natural language

With the built-in Coding Agent, turn a task description into an editable, debuggable, and saveable automation script so repetitive flows can be reused long-term.

Describe the goal

State the goal in natural language and the AI understands it.

AI generates the script

Produces a runnable script covering common actions.

Debug online

Trial-run and tweak instantly to validate the logic.

Save as a standard flow

Save as a template for long-term reuse and team sharing.

Capabilities

Powerful product capabilities

01

Universal AI tool integration

VMOS Edge works with Claude Code, Codex, Cursor, Gemini and more, and exposes virtual-phone control via MCP and Skills, helping developers and teams plug mobile operations into their existing AI workflows.

Connect directly with mainstream AI toolsMCP Guide Skills Install Guide
AI Tool Ecosystem · Topology
Plug & play
CLI
Claude CodeGemini CLICodex CLIQwen CodeKimi CLI+N
IDE / Editor
CursorVS CodeWindsurfClineZedTrae+N
Desktop
ChatGPTGoose AIOpenCodeFactory Droid+N
Connect via MCP / Skills
Connection
MCP
Standard protocol
Capability
Skills
Packaged controls
CORE
VMOS Edge · Cloud-phone control
02

AI Agents operate the virtual phone

Describe the goal and the AI recognizes the screen to search, tap, type, and verify, returning results in real time.

Say the goal directly AI acts live Results return live
AI Agent · Task Loop
No scripting needed
USER TASK
"Open Douyin, search 'camping', and like the first 3 videos"
01
Read the screen
Visually understand the page
02
Tap / type
Operate controls precisely
03
Verify result
Check if the goal is met
04
Return live
Sync results to the caller
LOOP
VisionDecideVerifyUntil goal met
03

AI-generated reusable workflows

The built-in Coding Agent generates automation scripts from natural-language tasks, with editing, debugging, saving, and batch execution — turning repetitive work into long-lived reusable workflows.

Natural-language chat AI-generated scripts Debug and reuse
AI Workflow · Script Generation
Coding Agent
task.js
async function run(vm) {
  await vm.launch("com.xingin.xhs");
  await vm.tap("Search");
  await vm.input("camping gear");
  return await vm.screenshot();
}
01
Natural-language chat
Describe / refine intent
02
Coding Agent generates script
Editable, debuggable
03
Save & reuse / batch run
Build once, reuse long-term
Generate
AI chat
Debug
Live run
Execute
Batch
04

One device, many independent cloud phones

A single VMOS L1 runs many Android VM instances at once, each with its own system, storage, and network — ideal for parallel tasks, isolation, and batch management.

VMOS L1 · Isolation Architecture
12 in parallel
Host · RK3588S
8C / 32G / SSD
↓ Split by in-house virtualization engine ↓
VM-01
Android 10
VM-02
Android 11
VM-03
Android 12
VM-04
Android 13
VM-05
Android 14
VM-06
Android 15
VM-07
Android 16
VM-08
Android 10
VM-09
Android 11
VM-10
Android 12
VM-11
Android 13
VM-12
Android 14
Own OS
Android 10–16
Own storage
SSD partition
Own network
VLAN / proxy
05

Centralized management on PC

Manage devices and VMs from the desktop client with live screencast, file transfer, ADB debugging, and batch operations for efficient, controllable multi-device management.

VMOS Edge · Console
Overview
Devices
Screencast
ADB
Monitor
Online
248
+12
Running VMs
1,824
97%
API calls/s
3.6k
Devices
Batch
L1-001Running12 VMCast
L1-002Running12 VMCast
K1-014Casting8 VMCast
ARM-S1Running128 VMCast
Live screencast
●REC
●REC
●REC
●REC
06

Programmable automation control

REST API and a JavaScript SDK support VM start/stop, file push, screenshot, touch injection, and shell execution to build integrable, scalable mobile automation systems.

Full developer docs and API referenceDeveloper Guide
REST API + JS SDK
192.168.1.10:8088
Script / system
SDK · HTTP
VMOS Edge API
REST · live
Cloud-phone instance
Per instance
POST/activity/startLaunch app
POST/input/clickInject tap
POST/shell/execRun command
POST/file/pushPush file
GET/screenshotScreenshot
Start / StopTap / TypeShellScreenshotFile pushADB
Hardware

Multiple devices, flexible deployment

All computing happens locally; your data never leaves the box, meeting different compute needs.

VMOS L1 Local Box - 1
Desktop box

VMOS L1 Local Box

8-core chip, 32 GB RAM + 32 GB storage, up to 12 Android phones

Form
Desktop box
CPU
RK3588S, octa-core A76×4 + A55×4, 2.4 GHz
Memory
32 GB LPDDR
Storage
32 GB eMMC + M.2 NVMe SSD slot
Network
Gigabit Ethernet
System
Android 10-16
Dimensions
117 × 111 × 38 mm
Buy Now
VMOS MagicBox K1 - 1
Edge computing box

VMOS MagicBox K1

8-core chip, 16 GB RAM + 32 GB storage, up to 12 Android phones, Google CTS certified

Form
Edge computing box
CPU
RK3588, octa-core A76×4 + A55×4, 2.6/2.0 GHz, 6 TOPS AI
Memory
16/24/32 GB LPDDR5
Storage
32 GB eMMC + M.2 NVMe SSD slot (up to 4 TB)
Network
Gigabit Ethernet
System
Android 10-16
Dimensions
130 × 120 × 40 mm
Buy Now
ARM Server (16-Node) - 1
Standard 2U rack

ARM Server (16-Node)

2U rack with 16 RK3588S nodes, up to 128 Android phones

Form
Standard 2U rack
Max Nodes
16 RK3588S nodes
Processors
16 ARM processors
CPU Config
Octa-core A76×4 + A55×4, 2.4 GHz
Memory
16-32 GB LPDDR4X per node
Storage
32 GB eMMC, expandable local storage
Network
20 Gbps (2×SFP+ 10G + 1G mgmt)
System
Android 10-16
Buy Now
ARM Server (128 ARM) - 1
2U rack + blade architecture

ARM Server (128 ARM)

2U blade server with 128 RK3588S nodes, up to 1,024 Android phones

Form
2U rack + blade architecture
Max Scale
Up to 16 blade units
Processors
Up to 128 ARM processors
CPU Config
Octa-core A76×4 + A55×4, 2.4 GHz
Memory
16-32 GB LPDDR4X per node
Storage
32 GB eMMC, expandable local storage
Network
40 Gbps, VLAN, port bonding, L3 switching
System
Android 10-16
Buy Now
Use Cases

Versatile Business Scenarios

From development and testing to commercial operations — VMOS Edge scales to your needs.

Multi-Version App Compatibility Testing
01

Multi-Version App Compatibility Testing

Run parallel tests across multiple Android versions to quickly surface compatibility issues and speed up development.

Automation Scripts & Batch Tasks
02

Automation Scripts & Batch Tasks

Automate repetitive tasks via API, reduce labor costs, and execute bulk operations at scale.

Isolated Multi-Account Operations
03

Isolated Multi-Account Operations

Each VM runs independently, providing full account isolation for safe and reliable multi-account management.

Teaching & Demo Screen Mirroring
04

Teaching & Demo Screen Mirroring

Low-latency mirroring for classroom instruction, product demos, and live operation walkthroughs.

Get Started

AI virtual-phone automation starts with VMOS Edge

Free Download