
Kun
Local-first personal AI workbench with desktop and terminal interfaces
- Platforms
- macOS, Windows, Linux
- GitHub stars
- 6.4K
- Last commit
- 3 days ago
- Language
- TypeScript
- Listing updated
- today
Similar agents
About Kun
Kun is a hosted agent for personal assistants. It is available on macOS, Windows and Linux.
Code, Work and chat share one local runtime, and the desktop GUI and terminal TUI share threads, model connections, approvals and background tasks. Sessions, memory and logs stay on your machine; you connect your own model through an API key or existing subscription.
Code combines projects, terminal, Git and diff with a Design canvas for comparing screens before implementation. Work handles documents and a paper library. In chat you can talk to a personal agent, Codex, Claude Code or OpenCode, or put several in a group to divide tasks.
It supports OpenAI and Anthropic compatible APIs and local models, MCP and skills. The agent asks for confirmation before acting. License is not stated.
What it does
- Code workspace
- Projects, terminal, Git and diffs in one place.
- Design canvas
- Compare screens and states before handing to an agent.
- Group agent chat
- Personal agent, Codex, Claude Code and OpenCode can collaborate.
- Paper library
- Import, read and annotate papers.
- Reports and artifacts
- Turn data into charts and saved reports.
- Local data
- Sessions and memory stay on your machine.
Good for
- Coordinate several coding agents in one chat
- Write and revise documents with selection-based edits
- Manage a paper reading list
- Schedule reminders and recurring tasks
Kun FAQ
Which platforms are supported?
macOS, Windows and Linux installers.
Which models can I connect?
OpenAI or Anthropic compatible APIs and local models.
Does it act without asking?
The README says the agent asks for confirmation first.
What are the best alternatives to Kun?
The closest alternatives to Kun are LobsterAI, CowAgent, QwenPaw, AionUi and openagent.
Where this comes from
- Description: written from the project's README and the project's website.



