Connect your agent to a desktop
Every way an agent reaches a Kubi screen: an agent app through our hosted MCP server, with one URL and no code, or your own code through the TypeScript SDK or the API. All of them reach the same screens, with the same permissions and the same history.
Anthropic
Claude Code
a coding agent in your terminal. It can check its own work in a real browser.
Anthropic
claude.ai and Claude Desktop
the Claude app, in a browser or on your desktop. No code, and nothing to install.
OpenAI
Codex
OpenAI's coding agent, in the terminal or the cloud. A machine that is not wiped between runs.
Anysphere
Cursor
the AI editor. The loop closes inside the editor.
Microsoft
VS Code
the editor, with its agent mode. Nothing to install beside the editor.
the SDK and the API
Your own code
TypeScript, or any language over HTTP. A screenshot you can act against.
Which way in?
All three reach the same screens, the same permissions and the same record.
| You have | Use |
|---|---|
| An agent app: Claude Code, claude.ai, Codex, Cursor, VS Code | The hosted MCP server |
| Your own code in TypeScript | The SDK, @trykubi/sdk |
| Your own code in any other language | The HTTP API |
| An agent that should keep running with your laptop shut | Install it on the screen itself |
An agent app we have not listed works too if it speaks MCP: the address is the whole setup. Agents we have not run ourselves, OpenClaw and Hermes Agent among them, get a page here once we have.










