AionUi is a free, open-source Cowork application where AI agents work alongside you directly on your computer—reading files, writing code, browsing the web, and executing automated tasks. You can see every step the agent takes and always stay in control.
[Purpose] AionUi is positioned as a multi-agent collaboration workbench. It is not just a chat client, but a platform that lets AI agents work side by side with you on your computer, suitable for scenarios such as file operations, document office work, data processing, scheduled automation, and remote control.
[Built-in agents and zero configuration] AionUi comes with a complete AI agent engine and is ready to use after installation: no need to install CLI tools, no complex configuration, and you can get started by pasting any API Key. It has complete agent capabilities such as file reading and writing, web search, image generation, and MCP (Model Context Protocol) tools, and includes 21 professional assistants that are ready to use.
[Office assistants: PPT, Word, and Excel] The built-in OfficeCLI supports quickly turning requests into deliverable documents: it can output editable Morph PPT (.pptx, with Morph animated transitions and coherent narrative pacing), editable Word (.docx, suitable for academic writing and production-grade document editing), and usable Excel (.xlsx/.xlsm/.csv, for analysis, automatic formatting, and chart generation), with final outputs that can be directly edited and reused.
[Multi-agent mode] If you already use Claude Code, Codex, Hermes Agent, OpenClaw, etc., AionUi can automatically detect them and let you collaborate with them as well as the built-in agents. Supported agents include: built-in agent (powered by the embedded aionrs engine), Claude Code, Codex, Qwen Code, Gemini CLI, Goose, OpenClaw, Augment Code, CodeBuddy, Kimi CLI, OpenCode, Factory Droid, GitHub Copilot, Qoder, Mistral Vibe, Nanobot, Snow, Hermes, Cursor Agent, Pi, MiMo Code, omp, Antigravity, etc. It provides automatic detection, a unified interface, parallel sessions, unified MCP management, and YOLO/fully automatic mode (The available modes and permission behaviors depend on the selected agent).
[Team Mode] Multiple AI agents can be formed into a team to collaborate: the Leader agent receives instructions, breaks them down into subtasks, and dispatches them to Teammate agents through the built-in Team MCP Server. Teammates execute in parallel, share results through an asynchronous mailbox, and write to a shared task board. Features include: parallel execution, Leader orchestration (supporting backends such as Claude Code, Codex, Hermes Agent, Gemini, Snow CLI, and Aion CLI), isolated team workspaces (sharing the same folder, with independent permission dialogs and sidebar badges), a shared workspace (all agents read from and write to the same folder, and the file panel is always visible), dynamic scaling (Teammates can be added or removed during runtime, silent agents are automatically escalated to failed and can be removed with one click), and fine-grained permissions and file sharing. External agents connect through ACP (Agent Client Protocol), and other ACP backends with mcpCapabilities.stdio are automatically supported.
[Any API Key, Full Agent Capabilities] Regardless of which model is used to drive it, agent capabilities are consistent (file reading and writing, web search, image generation, tool use). More than 30 AI platforms are supported, including official platforms (Gemini, Gemini Vertex AI, Anthropic Claude, OpenAI), cloud service providers (AWS Bedrock, New API), Chinese platforms (Dashscope/Qwen, Dashscope Coding Plan, Zhipu, Moonshot/Kimi, Qianfan, Hunyuan, 01.AI, ModelScope, InfiniAI, CTYun, StepFun, SiliconFlow-CN, PPIO), international platforms (DeepSeek, MiniMax, Novita, OpenRouter, SiliconFlow, xAI, Ark Volcengine, Poe), and local models (Ollama and LM Studio configured with local API endpoints through a custom platform). NewAPI gateway services are also supported, allowing flexible switching between different models in the same interface.
[Extensible Assistants and Skills] An extensible assistant system is provided: 21 built-in professional assistants, as well as a three-tier skill system (built-in skills, custom skills, and extension skills).. Built-in assistants include: Cowork (autonomous task execution), PPT Creator / Morph PPT / Morph PPT 3D, Pitch Deck Creator (investor pitch), Dashboard Creator (data dashboard), Word Creator, Word Form Creator, Excel Creator, Academic Paper Writer (academic papers), Financial Model Creator (financial models), 3D Game (single-file 3D games), UI/UX Pro Max (57 styles, 95 color schemes), Planning with Files (file-based persistent planning), HUMAN 3.0 Coach, Social Job Publisher, moltbook (zero-deployment agent social), Beautiful Mermaid (flowcharts, etc.), OpenClaw Setup, Story Roleplay (SillyTavern-compatible), AionUi Butler (product guidance and troubleshooting). Custom skills can be imported or created in Settings → Skills and mounted to assistants; the skill indicator in the conversation header shows the skills active in the current session, with search and exclusion supported. Built-in skills include pptx, docx, pdf, xlsx, mermaid, and more.
[Cowork Anywhere and Scheduled Tasks] WebUI mode supports access via browser on phones, tablets, or any computer, supports LAN, cross-network, and server deployment, and allows login via QR code or password. Chat platform integration supports Telegram, Feishu (Lark), DingTalk (AI Card streaming output with automatic fallback), and personal WeChat accounts; WeCom, Slack, Discord, and others are coming soon; the configuration entry is Settings → WebUI Settings → Channels, where you enter the Bot Token. Scheduled tasks support natural language descriptions and three scheduling modes (standard cron expressions with timezone support, fixed intervals, and one-time triggers); agents can also autonomously create scheduled tasks during conversations; execution mode can be set to continue in an existing conversation (retaining full context) or create a new session each time; tasks are bound to conversations and support create/modify/enable-disable/delete/view, automatically prevent system sleep during execution and detect missed triggers after wake-up, and each task can independently set the model, workspace directory, and reasoning effort. Examples: daily weather reports, weekly sales data summaries, monthly backup organization, custom reminders.
[Preview Panel and File Management] The preview panel supports instant viewing of files generated by agents in more than 10 formats: documents (PDF, Word, Excel, PowerPoint), code (more than 30 programming languages), markup (Markdown, HTML), images (PNG, JPG, GIF, SVG, WebP, BMP, ICO, TIFF, AVIF), and Diff files; it automatically tracks file changes, supports real-time editing of Markdown, code, and HTML, and opens in multiple tabs. File management supports automatic categorization and organization, batch renaming, file merging, content-based classification, and automatic execution of file operations.
[Limitations] Information not explicitly provided in the materials, such as fees, pricing, and platform availability, is not yet available; YOLO/full-auto mode and the permission behavior of each agent depend on the selected agent, and specific differences must be based on actual use.
