AionUi is a free and open-source app that provides a clean graphical interface for command-line AI tools like Gemini CLI, Claude Code, Qwen Code, and more on macOS, Windows, or Linux. It automatically recognizes these tools for direct conversations; It supports local storage of multi-session chat records, intelligent organization of files, instant preview of more than 9 formats such as PDF and code, and can also generate and edit images, and provide web page access capabilities. Use it to quickly and securely get AI assistance in office, coding, data processing, and more without complex commands, saving time and improving efficiency, and data never leaves your device.
With the trend of AI tools gradually becoming command-line-based, more and more powerful model tools exist in the form of CLI (command line), such as Gemini CLI, Claude Code, Qwen Code, etc. These tools are fully functional and scalable, but for users who are not used to frequent command input, the barrier to interaction still exists.
AionUi is not positioned to re-implement model capabilities but to provide a unified graphical interface encapsulation layer for these command-line AI tools.
Visualization shell for CLI
AionUi is an open-source desktop app that runs in a local environment. Its core idea is:
Without changing the logic of the original CLI tools, call the local installed AI command-line program through the graphical interface.
In other words:
- Model capabilities still come from tools like Gemini CLI, Claude Code, and more
- AionUi is responsible for the conversation interface, session management, and presentation layer
This structure is similar to how a “front-end shell + local AI engine” is combined.
Basic functional structure
From a technical perspective, AionUi typically includes the following aspects:
Multi-session management
Support saving chat history locally and switching between different tasks. Data is stored in local files or local databases, not synchronized in the cloud.
command to execute the package
By calling the CLI tool installed on the system, you can:
- Send a prompt to the model
- Receive streaming output
- Show results
This is essentially an encapsulation of the child process.
Documents and results presentation
For the content generated by the model, the interface layer supports:
- Markdown rendering
- Code highlighting
- Image shows
- Preview of common files (specific format subject to project description)
The goal is to improve readability, not change the model logic.
Local operation and data path
AionUi itself runs on the user’s local device and does not act as a cloud broker server.
But two situations need to be distinguished:
- If the CLI calls a local model, such as a large on-premises model, the data does not leave the device
- If the CLI calls a cloud API (such as Anthropic or Google Cloud Model), the data will still be sent to the corresponding model service provider
Therefore, AionUi only guarantees that “the UI does not upload data” without changing the network behavior of the model.
Scene
AionUi is better suited for the following types of users:
- Tools like Gemini CLI / Claude Code are already being used
- Want to reduce command line interaction
- Multi-session management and results visualization are required
- You want to unify the management of different AI tools in a desktop environment
It is not:
- New model platform
- New Agent framework
- or multi-agent systems
Instead, it is an interactive layer solution.
Relationship to AI workflows
In the overall AI application architecture, AionUi belongs to the presentation layer:
模型 / CLI 工具
↑
AionUi（桌面 GUI）
↑
用户交互It can be integrated with local development processes, coding workflows, and data processing scripts, but it is not responsible for model training, inference optimization, or tool orchestration logic.
Github：https://github.com/iOfficeAI/AionUi
Tubing: