Design Documentation¶

Technical architecture and design details for AutoMobile.

Overview¶

AutoMobile is a mobile UI automation framework built on the Model Context Protocol (MCP). It enables AI agents to interact with Android and iOS devices for testing, exploration, and automation.

flowchart TB
    subgraph Clients
        Agent["🤖 AI Agent"]
        IDE["💻 IDE Plugin"]
        CLI["🧪 Test Runner"]
    end

    subgraph MCP["MCP Server"]
        Transport["Transport Layer
(STDIO / HTTP)"]
        Registry["Tool & Resource
Registry"]

        subgraph Features["Feature Composition"]
            Actions["Actions
(tap, swipe, input)"]
            Observe["Observation
(hierarchy, screenshot)"]
            AppMgmt["App Management
(launch, terminate)"]
            DeviceMgmt["Device Management
(list, start, kill)"]
        end
    end

    subgraph External["External Interfaces"]
        WS["📡 WebSocket
Clients"]
        Socket["🔌 Unix Socket
Commands"]
        CLITools["🛠️ CLI Tools
(adb, xcrun)"]
    end

    subgraph Devices["Devices"]
        Android["📱 Android"]
        iOS["📱 iOS"]
    end

    Agent --> Transport
    IDE --> Transport
    CLI --> Transport
    Transport --> Registry
    Registry --> Features

    Actions --> WS
    Observe --> WS
    AppMgmt --> CLITools
    DeviceMgmt --> CLITools
    DeviceMgmt --> Socket

    WS --> Android
    WS --> iOS
    CLITools --> Android
    CLITools --> iOS

    classDef client fill:#FF3300,stroke-width:0px,color:white;
    classDef mcp fill:#525FE1,stroke-width:0px,color:white;
    classDef external fill:#00AA55,stroke-width:0px,color:white;
    classDef device fill:#666666,stroke-width:0px,color:white;

    class Agent,IDE,CLI client;
    class Transport,Registry,Actions,Observe,AppMgmt,DeviceMgmt mcp;
    class WS,Socket,CLITools external;
    class Android,iOS device;

Core Components¶

Design Principles¶

High quality, performant, and accessible workflows are necessary to creating good software.

1. Fast & Accurate - Mobile UX can change instantly so slow observations translate to lower quality analysis.
2. Reliable Element Search - Multiple strategies with integrated vision fallback.
3. Autonomous Operation - AI agents explore without human guidance or intervention.
4. CI/CD Ready - Built for automated testing pipelines, dogfooding the tool to test itself.
5. Accessible - You shouldn’t need to be a build engineer or AI expert to use AutoMobile.

Platform Support¶

Android¶

iOS (Planned)¶