nanoMuse: Open-source AI agent for phones and computers released

Introduction of nanoMuse

nanoMuse is a newly released open-source personal AI agent designed for deployment across various personal devices, including phones and computers. The project emphasizes user control and privacy, allowing for self-hosting of its components.

Core Functionality and Design

Unlike conversational AI models that primarily answer questions, nanoMuse is built to perform actions. It maintains continuity across devices, remembering past interactions and operating in the background even when the application is closed. The agent is designed to seek user confirmation before executing irreversible actions.

Technical Architecture and Open Source Nature

The entire nanoMuse ecosystem, including the phone app, desktop app, web console, and the relay that connects them, is available under the GPL-3.0-or-later license. This open-source approach allows users to inspect, modify, and deploy the system themselves. The project provides options for users to self-host the relay server, ensuring data privacy by keeping information within their own infrastructure, or to utilize a community relay with a free usage allowance.

Cross-Device Operation and Privacy

Each device runs its own agent instance, with the phone app using Alpine Linux under proot and the desktop app integrating DeepSeek Harness with a Python runtime. When signed in, these agents communicate via the relay, enabling shared conversations across devices. While conversation text travels through the relay, files and screenshots remain on the device where they were created, addressing privacy concerns.