Run nanoMuse, an open-source personal agent, across phone, desktop and web
The nanoMuse project is an open-source personal agent that runs on Android, iOS, Windows, macOS, Linux and the browser under one account. It can operate apps through their screens, keep working in the background and ask before irreversible actions.
The problem
Most assistants answer questions, run in a single cloud VM, or live only on the one device where they are installed. Apps without an API are hard for them to reach, and users have no easy way to share one agent across their phone and computer.
How they did it
- Install the app for your platform (Android APK, iOS TestFlight beta, macOS, Windows, Linux) or try the browser demo.
- Sign in with an e-mail or a mainland-China phone number to get a free allowance of model use on the community relay.
- Optionally add your own key from one of twenty providers, or use an existing ChatGPT, Claude or Kimi plan.
- Turn on Hands so the agent can operate phone apps and computer windows through their screens (Android and computers only; per the README, iOS does not allow it).
- Optionally self-host: run scripts/self-host.sh for your own relay, or docker compose up -d app for the web app.
From the official docs
bash scripts/self-host.sh --local
Results as reported by the project's author
- The article gives no performance metrics; it is a project README.
- Per the README, the phone, desktop and web share one account and show the same conversations.
- Per the README, a task said on the phone can run on a PC, with approvals returning to the device in hand.
- As of Oct 11, 2026, the repository shows 855 stars and 92 forks.
These figures are as reported by the project's author in nanoMuse GitHub README ↗. AgentGid did not measure or independently verify them.
This suits an individual or a small technical team comfortable with APKs, TestFlight betas and optionally Docker or shell scripts for self-hosting. The README gives no performance metrics, and with 855 stars and 92 forks (Oct 11, 2026) it is still a young project, so expect rough edges before letting Hands operate apps. If you want a more established open-source option, OpenClaw (Free (OSS)) ranks first in AgentGid's General category.
The agent used here
Similar use cases
How Rokt built 135 internal apps in a 24-hour hackathon with Replit
Rokt, an ecommerce technology company, let technical and non-technical staff build internal tools with Replit. During a global hackathon, employees produced 135 applications in a …
As reported by Replit: 135 applications were built within 24 hours during the hackathon.
How Incarna let agents pay BlockRun per inference with AgentCore payments
Incarna, which gives AI agents a persistent identity, used Amazon Bedrock AgentCore payments so its agents can pay BlockRun for model inference one request at a time over x402. Th…
As reported by AWS, the Incarna team finished the full integration in three days (one to build, two to test), against two to three months originally …
Search and read your WhatsApp messages with a read-only MCP server on Cloudflare Workers
An open-source project deploys a small server to your own Cloudflare account and links it to your phone as a WhatsApp linked device. Any MCP client can then search and read your c…
The agent gets read-only tools to list chats, read a chat, run full-text search across all messages, and trigger a sync. (as reported by the project's author)