A novel framework called PalmClaw enables the deployment of native on-device agents on mobile phones, allowing Large Language Model (LLM) agents to execute tasks directly on these devices. This development is significant because it brings the capabilities of LLM agents to a widely accessible and highly personal environment, where they can interact with user data and sensors. By running natively on mobile devices, PalmClaw-based agents can perform tasks without relying on remote servers or desktops, which enhances their responsiveness and reduces latency. The framework's ability to support multi-step tasks and tool integration1 makes it a powerful tool for automation and decision-making on mobile devices. This matters to practitioners because it has implications for the development of more sophisticated and autonomous mobile applications, which can leverage the unique capabilities of mobile devices to improve user experience and productivity.