nano-muse
nanoMuse is an open-source personal agent for every device a person owns. It acts through a phone, computer and web console, shares one conversation across devices, remembers the user in editable files, and asks for approval before irreversible actions.
At a Glance
- Individual users who want a personal agent across their own devices
- Open-source developers and contributors
- Privacy-conscious users and self-hosters
- Researchers and developers working on personal agents, computer use and embodied/screen agents
AI Tools by nano-muse
(1)
nanoMuse
Open Source Cross Device Agent
Discussions
No discussions yet
Be the first to start a discussion about nano-muse
Latest News
nanoMuse 1.0.0 ‘Keel’ is the first stable release across phone, desktop, browser and relay.
nanoMuse paper appeared on Hugging Face Daily Papers; the project README reported a third-place ranking for October 8.
The nanoMuse paper became available on arXiv: ‘nanoMuse: An Open-Source Personal Agent for Every Device You Own.’
Project homepage announced release 0.1.41 ‘Choice’.
Products & Services
Android 8.0+ arm64 client with an on-device agent, shell, browser, MCP and skills, editable Markdown memory, screen-based Hands, approvals and cross-device conversation sync.
Desktop client for Windows, macOS and Linux. It runs the agent locally, can use computer windows as Hands through screen interaction, and connects to the shared hub and relay.
iOS/iPadOS beta distributed through TestFlight, sharing the account and conversations with the other clients.
Browser client/demo at demo.nanomuse.dev, with a web app for the same agent and account.
Market Position
nanoMuse positions itself as an open-source, self-hostable counterpart to Meta’s closed Muse, differentiated by running across the user’s phone, computer and web rather than only in a vendor cloud, supporting user-selected models, readable memory and an approval Sentinel. It is also broader than OpenMinis by spanning multiple device types and a relay, while building on OpenMinis for the phone line.
Leadership
Founders
Guangyi Liu
PhD student in the College of Control Science and Engineering at Zhejiang University; his stated research interests are language agents, digital automation and artificial intelligence. He is the primary GitHub maintainer under lgy0404.
Yong Liu
Professor at Zhejiang University’s Institute of Cyber-Systems and Control. He holds a PhD and BSc in Computer Science from Zhejiang University and researches intelligent robot systems, robot perception and vision, deep learning, big-data analysis and multi-sensor fusion.
Jiangning Zhang
Zhejiang University researcher/PhD student associated with the Institute of Cyber-Systems and Control; his listed interests include low-level computer vision, generative adversarial networks and neural architecture design. His prior research includes lightweight vision models such as EMOv2.
Founding Story
The project’s paper describes nanoMuse as an open-source counterpart to the closed, cloud-based personal-agent concept exemplified by Meta’s Muse. The initial vision was one agent across a person’s devices, with a shared conversation, user-readable memory, screen-based ‘hands,’ self-hostable relay, model choice and an approval Sentinel for actions the user could not undo.
Business Model
Revenue Model
The project describes itself as free, open-source and non-profit. A community relay provides a free allowance of model use funded by the developer; users can instead supply their own provider key, and the relay/runtime can be self-hosted.
Pricing Tiers
Sign in to the community relay to receive a free allowance of model use and device synchronization.
With a key of the user’s own, the runtime needs no account; users can also run the relay on their own server.
Target Markets
- Individual users who want a personal agent across their own devices
- Open-source developers and contributors
- Privacy-conscious users and self-hosters
- Researchers and developers working on personal agents, computer use and embodied/screen agents
- Cross-device personal task execution
- Automating desktop and mobile apps through their screens
- Browser, shell, MCP and skill-based workflows
- Scheduled personal routines, goals and daily feeds
- Persistent personal memory and context across weeks
- Privacy-conscious or self-hosted personal-agent deployments