AionUi Cowork
Run built-in or CLI agents in one desktop and WebUI workspace for file work, office outputs, and scheduled automation.
- Source repo
- iOfficeAI/AionUi
- Stars
- ★ 34k
- Last updated
- 1d ago
- License
- Apache-2.0
- Primary language
- TypeScript
- FA score
- 43/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code · OpenAI API · Claude API
- Cost
- Free software; you pay for model usage
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- A developer who uses both Codex and Claude Code on a local project can launch separate, context-isolated sessions from one interface.
- Not a fit if
- Developers who want a pure CLI or CI pipeline tool
- Organizations unwilling to grant local file access and unattended execution
- Source review
- 43/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
AionUi is a cross-platform Cowork application for local computers, with a built-in agent and a unified interface for installed CLI agents. After configuring a supported provider's API key, its built-in agent can read and write files, search the web, generate images, and use MCP tools. It auto-detects CLI agents including Claude Code, Codex, and Gemini, and can run parallel sessions with separate contexts or coordinate them as Leader and Teammate agents. The product offers a desktop GUI and a WebUI that can run as a standalone HTTP server for browser, phone, or remote access. Its FAQ says application data is stored in a local SQLite database, while OfficeCLI supports editable presentation, Word, and spreadsheet outputs.
A user selects the built-in agent or a detected CLI agent in AionUi and submits work in a conversation; the agent can read and write local files, search the web, generate or edit images, and call MCP (Model Context Protocol) tools. Multi-agent mode places sessions from different CLI tools in one interface, while Team Mode has a Leader split work and delegate it through the built-in Team MCP Server to Teammates, with results shared through an async mailbox and task board. Built-in skills including pptx, docx, xlsx, pdf, and mermaid support file-based workflows; OfficeCLI produces editable .pptx, .docx, .xlsx, .xlsm, or .csv files. Scheduled tasks can use five-field cron, fixed intervals, or one-time triggers, either continuing an existing conversation or creating a new one each run.
- A developer who uses both Codex and Claude Code on a local project can launch separate, context-isolated sessions from one interface.
- An operations analyst preparing weekly sales reporting can generate or refresh
.xlsx/.csvdata, formatting, and charts, then schedule the workflow. - An office user who needs editable slide decks, papers, or contract forms can use Morph PPT, Word Creator, or Word Form Creator and retain
.pptxand.docxdeliverables. - A person who wants to check or continue work running on an office computer from a phone can use the browser-accessible WebUI.
- A team dividing a complex request across models can use Team Mode for a Leader to delegate, track, and collect parallel Teammate work.
How do you install or deploy this agent?
Download the platform-specific installer for macOS, Windows, or Linux from GitHub Releases. On macOS, Homebrew installation is also documented: brew install aionui. Documented requirements are macOS 10.15+, Windows 10+, or Ubuntu 18.04+/Debian 10+/Fedora 32+, with 4GB+ memory recommended and 500MB+ available storage. Open AionUi after installation and enter credentials or an API key for a supported platform such as Gemini, OpenAI, Anthropic, AWS Bedrock, Ollama/LM Studio, or NewAPI; the built-in agent works without a separately installed CLI.
How do you use this agent?
For a first session, enter the selected provider's API key in AionUi, create a Cowork conversation, and send the built-in agent a task such as organizing files or generating a document. If Claude Code, Codex, or another supported CLI is already installed on the computer, AionUi auto-detects it; select that agent to run it from the unified interface. For remote access, open Settings → WebUI Settings → Channel and configure WebUI or a chat channel; Telegram, Lark, DingTalk, and WeChat require a Bot Token. For automation, create a conversation-bound scheduled task and choose cron, a fixed interval, or a one-time trigger.
What are this agent's strengths and limitations?
- The built-in agent does not require a CLI installation first; a supported provider API key is sufficient to start.
- It centralizes installed CLI agents such as Codex and Claude Code alongside the built-in agent, with parallel isolated sessions and Team Mode coordination.
- MCP tools can be configured centrally and synchronized to agents instead of being configured one CLI at a time.
- OfficeCLI and the built-in office skills target editable PPTX, DOCX, and XLSX/CSV deliverables rather than chat-only results.
- Cloud-model use still requires the user's own API key and incurs usage charges from the selected provider.
- CLI auto-detection only integrates tools already installed locally; obtaining those agents' capabilities still requires their separate installation and setup.
- The product includes local file read/write and autonomous execution; YOLO Mode and Full-Auto Mode bypass manual approvals, which raises permission-review risk.
- Local development is documented as spanning the AionUi Electron frontend and a separate AionCore backend repository, rather than a single-repository setup.
How does this agent compare with similar options?
The repository compares AionUi with Claude Cowork: it positions AionUi as a multi-model Cowork application for macOS, Windows, and Linux with desktop GUI, WebUI, scheduled tasks, and multiple chat channels, while describing Claude Cowork as macOS-only and Claude-only.
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| AionUi Cowork This agent | 43 · Major gaps | Desktop appFree + model costs | ★ 34k | 1d ago | TypeScript | Codex · Claude Code · OpenAI API · Claude API |
| somora | 69 · Some gaps | CLIFree | ★ 35 | 10d ago | TypeScript | ChatGPT · Codex · Claude Code |
| QwenPaw Personal Assistant | 64 · Some gaps | CLIFree + model costs | ★ 37k | 1d ago | TypeScript | OpenAI API · Claude API |
| NextClaw | 58 · Major gaps | Desktop appFree + model costs | ★ 263 | 1d ago | TypeScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence: README claims permission confirmation dialogs (each agent has its own permission dialog), but no implementation details; YOLO/full-auto mode mentioned without permission boundaries; data flow transparency only mentions 'see everything the agent does' without specifics; sensitive data handling not mentioned; dependency security has resolutions and patchedDependencies but no vulnerability scan evidence; external effects include file operations, network access, but default permissions not specified; rollback mechanism not mentioned; source attribution has author info but publisher unverified. Deductions: permissions and confirmation mechanisms only claimed, no code evidence; data flow and sensitive data handling lack specifics; dependency security only partial measures; external effects default permissions unclear; rollback missing; publisher identity unknown.
Evidence: README features align with package.json dependencies and scripts, but placeholder test (equivalence.test.ts is no-op) indicates self-consistency issues; many dependencies with pinned versions but no availability guarantee; failure messages not mentioned. Deductions: placeholder test shows insufficient test coverage; dependency availability unverified; failure messages missing.
Evidence: README clearly identifies target users (developers, enterprises) and use cases (office automation, remote access), audience and scenarios clear; capability boundaries described (supports multiple agents, platforms) but not explicit limits; trigger precision not detailed; environment fit has cross-platform support (macOS/Windows/Linux) and multiple deployment modes. Deductions: capability boundaries and trigger precision lack specifics; environment fit broad but no configuration details.
Evidence: README well-structured with quick navigation; install notes have download link and quick start; naming stable (AionUi); examples and FAQ via community links; known limitations not explicitly listed; license Apache-2.0; version in package.json but no changelog; maintenance responsibility has author info but not explicit. Deductions: known limitations missing; versioning changelog missing; maintenance responsibility unclear.
Evidence: Output usability has preview panel supporting multiple formats; marginal value high (free, open-source, multi-agent integration); cost-benefit good (free). Deductions: output usability relies on external tools (OfficeCLI), unverified; marginal value based on claims, unverified; cost-benefit based on free, but maintenance cost not considered.
Evidence: README claims partially linked (e.g., AionCore directory), but no concrete evidence; cross-source verification limited (only README and package.json); facts and inferences not clearly separated. Deductions: claims lack traceable evidence; cross-source verification limited; facts and inferences mixed.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Publisher identity unverified, proceed with caution.
- Permission and confirmation mechanisms only claimed, no code evidence, actual behavior may differ.
- Placeholder test exists, test coverage may be insufficient.
- Many dependencies, pay attention to supply chain security.
- Known limitations not listed, potential hidden risks.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Highlights Explorer: tab-scoped refresh and Collapse All - The Files and Changes panel now has a single top-bar Refresh button that does the right thing per tab: on Files it re-mounts every root and revalidates project detail; on Changes it re-discovers repositories and re-pulls their status. - A git worktree created inside an already-open project now shows up on the Changes tab after a single refresh — no more switching projects away and back to make it appear. - The Refresh button tracks real in-flight state (busy/spinner) and ignores repeated clicks until the refresh completes. …View on GitHub ↗ -
Release notes excerpt
Highlights Clearer conversation notifications - Turn-completed notifications now include the conversation's name, so you can tell at a glance which conversation finished instead of seeing identical "finished responding" messages (long names are truncated). Sidebar tells "still working" from "needs your input" - When a conversation pauses waiting on you — a tool permission request or an agent question (AskUserQuestion) — the sidebar now shows a distinct attention icon instead of the generic generating spinner, across the ACP, Codex, and Aion CLI backends. …View on GitHub ↗ -
Release notes excerpt
Highlights Appearance: pick your own font family and weight - Settings → Appearance now lets you choose the typeface across four regions — app, chat, markdown and code — from a searchable list of fonts installed on your machine, plus a matching font-weight control (Light/Regular/Medium/Semibold/Bold, or system default) for each region. Archived conversations, fully reachable - The archived conversations list now pages in the rest of each group instead of silently truncating it — a "load more" button reveals older archived chats a page at a time. …View on GitHub ↗