Part 6: Complete Example - MobileAgent

Note: This comprehensive hands-on tutorial is currently under development. Check back soon for a complete MobileAgent implementation walkthrough.

What You'll Build

A fully functional MobileAgent that can:

Control Android/iOS devices
Perform UI automation
Execute touch gestures (tap, swipe, type)
Capture screenshots and UI hierarchy
Integrate with Galaxy orchestration

Planned Content

1. Platform-Specific Setup

Android

ADB (Android Debug Bridge) integration
UI Automator framework
Accessibility services

iOS

XCTest framework
Accessibility API
Instrument tools

2. Complete Implementation

Agent class
Processor and strategies
State manager
MCP server with mobile tools
Prompter for mobile UI

3. Advanced Features

Multi-device coordination
App-specific automation
Error recovery strategies
Performance optimization

Temporary Reference

For now, study the LinuxAgent implementation as a complete reference:

Key Files

Component	File Path
Agent Class	`ufo/agents/agent/customized_agent.py`
Processor	`ufo/agents/processors/customized/customized_agent_processor.py`
Strategies	`ufo/agents/processors/strategies/linux_agent_strategy.py`
States	`ufo/agents/states/linux_agent_state.py`
Prompter	`ufo/prompter/customized/linux_agent_prompter.py`
MCP Server	`ufo/client/mcp/http_servers/linux_mcp_server.py`

Quick Start Template

# Minimal MobileAgent structure (to be expanded)

@AgentRegistry.register(
    agent_name="MobileAgent",
    third_party=True,
    processor_cls=MobileAgentProcessor
)
class MobileAgent(CustomizedAgent):
    def __init__(self, name, main_prompt, example_prompt, platform="android"):
        super().__init__(name, main_prompt, example_prompt,
                         process_name=None, app_root_name=None, is_visual=True)
        self._platform = platform
        self._blackboard = Blackboard()
        self.set_state(self.default_state)

    @property
    def default_state(self):
        return ContinueMobileAgentState()

Agent Architecture - Architecture overview
Agent Types - Platform implementations
Linux Quick Start - LinuxAgent deployment

Previous: ← Part 5: Testing & Debugging
Back to Index: Tutorial Series