kevinzhow / Rungic

Rungic 
 AgentOS in your hand. 
 A budget-friendly Android phone, a full Linux desktop computer, and an assistant that does the work for you. 
## Contents
- Working together
- Just say it
- The assistant has its own screen
- A team of agents
- Proactive intelligence: useful suggestions, at your pace
- You stay in control
- A real computer in your pocket
- Still your Android phone
- How it works
- Performance
- What makes it Agent Ready
- Integrations
- Status
- Supported devices
- Skills
- Learn more
 “Make a little rocket in Blender and render it for me.” The whole task took about 2½ minutes. The animation plays faster. 
Rungic turns a compatible Android phone into a Linux desktop computer. It offers a budget-friendly way to use the hardware you already own. It opens the KDE Plasma desktop and runs desktop software such as Firefox, Blender, Krita and VS Code. Connect a TV to use it as a desktop PC.
Rungic includes an AI assistant that can see, speak and act. Tell it what you need. It opens apps and does the work while you watch.
**Rungic supports your choice of agent.** The bundled assistant is a working demonstration and the default implementation. It currently uses Codex. Other agents can use the desktop, phone interfaces and system services independently. Connecting another agent to the bundled assistant's voice, tasks and widgets requires an adapter. There is no universal one-click switch yet.
## Working together
**Once agents become lasting participants in our work, what should the relationship between people and computers look like?**
We imagine persistent workspaces where agents explore, build and learn. People can join the work, understand decisions and discoveries, and discuss what happens. They can take over a step, change direction and return the work to the agent. The aim is to free people's attention for understanding, judgment, learning and creation. Our Agent OS philosophy describes this direction.
## Just say it
Hold the Home button and ask:
> “Make a little rocket in Blender and render it for me.”
>
> “Install Krita for me.”
>
> “Put your screen on the TV.”
>
> “Send Mom a WeChat voice message: I'll be home for dinner.”
The assistant presents a plan first and speaks its progress updates. It opens apps, clicks buttons and types for you. When the task ends, its pictures and files appear in the conversation. You can open them there.
By default, Codex reads the screen and decides each desktop action using the model and sign-in selected for the Agent. In Settings → Desktop operation, you can choose the separate Luna API executor instead. Codex uses its existing ChatGPT or API-key sign-in; selecting a desktop executor does not change that sign-in. Voice services, speech synthesis and the optional Luna executor use the configured OpenAI API key. The implementation and acceptance record lists what has been tested on the development G100.
## The assistant has its own screen
The assistant works on its own desktop. You can continue using y