nanoMuse changelog: what's new each month
Every stable nanoMuse release summarised by month: the highlights, new features, improvements, fixes and anything you need to act on. 2 months covered; the current month updates daily.
October 2026
27 releases: 0.1.25 → 1.0.9
October brought nanoMuse from beta to version 1.0, introducing cross-device conversation sync, a unified desktop application built on DeepSeek Harness, and support for multiple AI providers alongside nanoMuse Cloud. The month focused on making the agent accessible across all platforms while letting users choose their own models.
- Token in URL query strings now rejected; tokens must be sent in the WebSocket frame or Authorization header.
- nanoMuse Cloud stopped funding image and video generation; picture and clip generation now requires your own provider.
- URL scheme changed from minis:// to nanomuse:// on iPhone; MCP callbacks now use nanomuse-mcp://.
Highlights
- Desktop application now unified on DeepSeek Harness with Feed, Ideas, Goals and Library rooms, plus live screen capture and control.
- Conversations sync across all your devices, with the ability to hand tasks between them using @device mentions.
- Support for multiple AI providers—ChatGPT, DeepSeek, OpenAI-compatible endpoints—alongside nanoMuse Cloud in a single settings interface.
- iPhone app arrived on TestFlight with full feature parity to Android and desktop, including voice input and scheduled tasks.
- Agent can now handle hands-off work: scheduled tasks ask for approval when you're not at the computer, and approvals sync across devices.
New
- One account conversation appears on every device; side chats stay local unless synced.
- Device-to-device task routing: send work to another device with @devicename in the message.
- Avatar studio lets you describe a character, pick from seven styles, and generate four options with motion clips.
- Memory tool for storing one-line facts the agent recalls in future conversations.
- Connectors to Notion, Linear, Atlassian, Asana, GitHub, Sentry, Vercel, Cloudflare, Supabase, Stripe, PayPal, Figma and Canva on mobile.
- Chat apps support: Feishu, DingTalk, WeChat Work and Telegram over their native connections.
- Scheduled tasks and routines that run on the desktop and ask for approval when needed.
- Custom MCP servers with stored credentials in the profile, shared across devices.
Improved
- Sign-in now shows which model answers during setup and remembers your choice across devices.
- Model picker on every client folds long provider lists and supports search.
- Data controls and permission approvals unified across phone, desktop and web.
- Desktop desktop layout mirrors the phone's: rail with agent face, chat column, and floating composer pill.
- Cloud allowance now shows across all clients with a heads-up at 80 percent.
- Task approval cards now float on the desktop while the agent works, with Allow once and Always in app options.
- Thinking model reasoning displays in real time, not only after the answer.
- Record of conversations improved: tracks who started each request and what provider answered.
Fixed
- Desktop app no longer crashes or hangs after updates; loopback connection request remembers its port.
- Android app no longer crashes during voice input when the microphone falls silent.
- Windows installer now includes the files needed for in-app updates.
- iPad composer returns and no longer disappears; chat scrolls correctly under header and composer.
- macOS Screen Recording and Accessibility permissions now requested through native system calls.
- Linux desktop opens again from the icon and no longer requires the menu bar to stay alive.
September 2026
23 releases: 0.1.1 → 0.1.24
September brought nanoMuse from a local Android app to a cross-platform ecosystem. The month saw the addition of web and desktop versions, cloud sign-in with free allowances, and new capabilities like hands-on automation, computer control, and avatar generation.
- Chat title no longer shown in top bar; rename moved to menu and show title setting removed.
- Sign-in now required on first run before accessing chat features.
- Local network computer pairing via script replaced with cloud account-based pairing.
Highlights
- Sign in with email or phone number to access nanoMuse across web, desktop and phone with shared conversations and devices.
- Hands on screen automation now works on Android, desktop and paired computers with visual feedback and approval checkpoints.
- Generate and animate custom avatars with AI image models, with cost preview before creation.
- Desktop and web versions of nanoMuse with the same design language and features as the phone app.
- Free lifetime allowance of ¥10 per account, with pay-as-you-go options when exhausted.
New
- Web app at nanomuse.cn/web requiring no installation, keeping state between visits.
- Desktop app for Windows, macOS and Linux with sidebar navigation and multi-window layout.
- Hands feature for screen automation with visual stage showing gestures and approvals for sensitive actions.
- Computer pairing through account sign-in instead of local network scripts.
- Avatar generation with four candidate styles to choose from before committing.
- Call-like voice and video interaction (later removed in version 0.1.22).
- Coding agent integration showing Cursor, Codex and Claude Code sessions from your computer.
- Invite codes for sharing nanoMuse accounts with others.
Improved
- Welcome screen now appears correctly on first run with setup steps and model selection.
- Main chat remembered between sessions and opened by default on app launch.
- Settings reorganized with Muse's design language: round back disc, centered title, white cards on grey.
- Home shell with four tabs (chat, ideas, goals, library) that maintain state when switching.
- Avatar mood animations with breathing at rest and movement while working or waiting.
- System font used throughout (MiSans on Xiaomi, Roboto elsewhere) for native platform feel.
- Image model selection unified with video models on one settings page.
- Approval system expanded to three tiers: install software (notify), delete and send (ask and remember).
Fixed
- Setup no longer disappeared on phones after home navigation replaced session list.
- First conversation now uses the model chosen in setup, not whichever was found first.
- Home composer and chat styling now matches Muse's design consistently.
- Avatar and picture generation on Alibaba Cloud Model Studio fixed for HTTP 404 errors.
- Default image model now matches recommended model in settings.
- Remote approvals from web app now reach correct device instead of local gate.
Summaries are written automatically from the official release notes (full changelog ↗); check the original notes before relying on a detail. nanoMuse: pricing, features and alternatives · All changelogs