OPPO's Multi-X team has unveiled X-OmniClaw, an open-source Android AI agent framework designed to operate primarily on local devices. This "edge-first" framework performs core control, perception, and execution processes on the device itself, utilizing cloud-based models only for complex reasoning tasks. X-OmniClaw aims to transform smartphones into continuous AI assistants, integrating camera, screen, and voice inputs to perform actions within real apps. The framework's architecture consists of three components: perception, execution, and memory. The perception layer uses a vision-language model to understand scenes, while the execution layer identifies interfaces for navigation and interaction. The memory layer retains contextual information across tasks. X-OmniClaw also features behavior cloning, allowing the system to replicate user-demonstrated paths using Android deeplink, and emphasizes long-term semantic memory for enhanced task continuity. The project, based on the HermesApp codebase, is available on GitHub, with OPPO planning further updates.