AionUi
iOfficeAI/AionUiReviewed by AI
Checked by RepoPicks Radar on
What is AionUi?
AionUi is a desktop program with a built-in AI agent that works on the files on your computer: it tidies folders, renames in batches, reads and writes Word, Excel and PowerPoint files, analyses spreadsheets, generates images and previews the results. You install it and paste an API key from a provider you choose (Gemini, OpenAI, Anthropic, DeepSeek, Kimi and others, or a local model through Ollama). It can also drive AI command-line tools you already have, run tasks on a schedule, and be reached from a phone through a browser or a Telegram bot.
How easy is it to install AionUi?
B · installer plus one simple step · Windows
- Windows B
- macOS B
- Linux B
Each platform has a ready installer (.exe for Windows, .dmg for Mac, .deb for Linux, plus `brew install aionui`), and the README's three steps are install, paste an API key, start working — an installer plus one simple step, so B everywhere. Minimum stated: Windows 10, macOS 10.15, Ubuntu 18.04, 4 GB of memory.
AionUi on GitHub
- Stars
- 34,194on Nov 10, 2026
- Downloads
- 933,354all releases
- Growth
- +160stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Nothing extra
- To build from source
- Node.js, Docker
- Disk space
- 452 MB–1.3 GB · from the installer
- Accounts or keys
- An AI model provider of your choice (Gemini, OpenAI, Anthropic, DeepSeek, Kimi, OpenRouter and 30+ more), Telegram, Lark, DingTalk or WeChat bot token (only for chatting with it from those apps)
- Cost
- Free, API costs extra · The app is free (Apache-2.0). You pay the AI provider for what you use; free tiers or a local model through Ollama or LM Studio can avoid that.
- License
- Apache-2.0 (Permissive)
- Can replace
- Claude Cowork, Manus
- Last change on GitHub
- Nov 5, 2026
AionUi risk signals (7)
- AttentionDependencies with known vulnerabilities
Pinned dependencies with known vulnerabilities (OSV): builder-util-runtime: GHSA-p2f4-r6v6-j797
automatic check
- Attentioncrash reporting despite local only claim
The README's Q&A says "All data is stored locally in a SQLite database. Nothing is uploaded to any server.", but package.json depends on "@sentry/electron": "^7.10.0", a service that sends crash and error reports to a remote server.
AI review
- Attentionagent can change your files unattended
The README's own table says "Yes — built-in agent with full file access" and lists "YOLO / Full-Auto Mode — supported agents can expose an unattended mode" plus "Cron — 24/7 unattended" scheduled tasks. A mistake while you are not watching can rename, move or overwrite your files.
AI review
- Attentionremote access opens a door
WebUI mode "Supports LAN, cross-network, and server deployment. QR code or password login", and the Dockerfile sets "ENV ALLOW_REMOTE=true" with port 3000 exposed. Whoever reaches that address and gets past the password can drive an agent that has access to your files.
AI review
- Infonewest installers not on github
The last GitHub release with installer files is v2.1.47-final (2026-08-05); the releases after it, up to v2.2.2 (2026-09-09), carry no files at all, while the code is still committed daily. New downloads have to come from the project's own site, aionui.com.
AI review
- Infocomponent downloaded after install
Dockerfile comment: "officecli (the Office preview component, auto-installed at runtime by the backend) is a .NET binary". Part of the Office features arrives from the internet after you install the app.
AI review
- Infopaid referral links in readme
The README promotes a Kimi partnership with affiliate links ("kimi.com/code?aff=aionui", "platform.kimi.ai?track_id=...&aff=aionui") and free paid plans "for our contributors", so some recommendations in it earn the project money.
AI review
Evidence is shown in its original language.
What Radar could not check about AionUi
- Whether the 21 built-in assistants produce documents as good as the animated demos
- What the Sentry crash reporting sends and whether it can be turned off
- How much you will actually spend on API calls for a day of agent work
- Whether every one of the 20+ listed command-line agents really is detected
- What is inside the installers on the project's website, since they are not on GitHub
- Whether the WebUI password login is strong enough to expose to the internet