Operit: a feature-rich Android AI agent with built-in Ubuntu 24 and autonomous UI automation

Operit: a feature-rich Android AI agent with built-in Ubuntu 24 and autonomous UI automation

What it solves

Operit AI is a comprehensive AI assistant application for Android that transforms a mobile device into a powerful, autonomous agent. It solves the limitation of mobile AI apps being simple chat interfaces by integrating deep system access, local model execution, and a vast ecosystem of tools and automation capabilities.

How it works

The app runs independently on Android (API 26+) and integrates several core technologies:

  • Local Inference: Supports MNN and llama.cpp (GGUF) for offline AI processing to ensure privacy.
  • System Integration: Built-in Ubuntu 24 environment allowing the execution of Linux commands, Python, and Node.js directly on the phone.
  • Automation: Uses a triple-channel automation system (Accessibility, ADB, and Root) combined with AutoGLM to perform UI-level actions and automatic clicking.
  • Extensibility: Implements the MCP (Model Context Protocol) and a Skill market for installing plugins and remote tools.
  • Memory & Context: Features an intelligent memory system that automatically categorizes and summarizes history to provide personalized services.

Who it’s for

  • Power Users: Android users who want a fully autonomous AI agent on their phone.
  • Developers: Those looking to run Linux environments, manage files via SSH/SFTP, or build web projects on mobile.
  • AI Enthusiasts: Users who prefer local LLMs for privacy or want to highly customizable AI personas (character cards).

Highlights

  • Full Linux Environment: Integrated Ubuntu 24 for complex automation and development tasks.
  • UI Automation Agent: AutoGLM integration for autonomous screen interaction and virtual screen support for root users.
  • Extensive Toolset: Over 40 built-in tools covering file systems, network requests, media processing, and deep search.
  • Advanced Persona Management: Support for character cards (SillyTavern/JSON) with independent conversation histories.
  • Hybrid Model Support: Works with cloud APIs (OpenAI, Claude, Gemini) and local GGUF models.

Sources