Oppo open-sources Android AI agent X-OmniClaw that uses your camera, screen, and voice without leaving the phone

Oppo's Multi-X team has released and open-sourced an AI agent named X-OmniClaw. Designed for Android devices, the agent's core feature is its ability to run directly on the mobile device locally. By integrating the camera, screen display, and voice interaction, it can process tasks in real time within applications. The key technical aspect is that the system does not rely on cloud-based phone mirroring but primarily uses the device's local sensors to perform operations. It only calls on cloud computing power for complex reasoning. The agent also has a "skill cloning" capability, which records a user's click operation paths and converts them into reusable skills. When executing similar tasks again, it can use DeepLink to jump directly to deep pages within the application, significantly improving operational efficiency. This approach combines multimodal perception, local real-time processing, and reusable skills, offering a new implementation path for on-device AI agents. The news was initially published by the tech media outlet The Decoder.