AionUi is a desktop app for macOS, Windows, and Linux where AI agents work on your computer alongside you: reading and writing files, writing code, browsing the web, and running multi-step tasks with your approval. It ships with a built-in agent engine (aionrs), so it works with only an API key and no separately installed CLI agent.
It also auto-detects CLI agents already on the machine and runs them in one interface with parallel, independently scoped sessions. Supported agents include Claude Code, Codex, Qwen Code, Gemini CLI, Goose, OpenClaw, Kimi CLI, OpenCode, GitHub Copilot, Nanobot, Hermes, Cursor Agent, and Antigravity. External agents connect through the Agent Client Protocol (ACP), and MCP servers are managed in one place and synced to each agent according to its capabilities.
In Team Mode, a leader agent splits an instruction into subtasks and delegates them to teammate agents through a built-in Team MCP server. Teammates run in parallel on their own models, share a workspace folder, exchange results through an async mailbox, and each get their own permission dialog.
The built-in agent runs on a key for Gemini, OpenAI, Anthropic, AWS Bedrock, or a NewAPI gateway, on local models through Ollama or LM Studio, or on one of 30+ platforms including DeepSeek, Qwen (Dashscope), Moonshot Kimi, Zhipu, MiniMax, OpenRouter, and xAI. It comes with 21 built-in assistants, including PPT, Word, and Excel creators that produce editable files through OfficeCLI, plus academic paper, financial model, and dashboard assistants.
Features
- Built-in agent: file read and write, web search, image generation, and MCP tools, with no separate CLI agent to install
- Multi-agent mode: auto-detects installed CLI agents and runs them in parallel sessions in one interface
- Team Mode: a leader agent delegates subtasks to teammate agents that share a workspace and a task board
- Model choice: 30+ cloud and local platforms, including Gemini, OpenAI, Anthropic, AWS Bedrock, DeepSeek, Qwen, Ollama, and LM Studio
- Office assistants: PPT (Morph transitions), Word, and Excel assistants that produce editable files through OfficeCLI
- Assistants and skills: 21 built-in assistants, custom assistants, and built-in, custom, and extension skills toggled per conversation
- Scheduled tasks: cron, fixed-interval, or one-time triggers bound to a conversation, which agents can also create themselves
- Remote access: a browser WebUI over LAN or the internet, plus Telegram, Lark, DingTalk, and WeChat bots
- Preview panel: live preview and editing of PDF, Office documents, code, Markdown, HTML, images, and diffs in tabs
- Permission control: per-agent approval dialogs, with an unattended mode on agents that support it
