Read, modify, and interpret screen recordings in the terminal with an AI coding agent
Once you enter a project folder and type the kimi command, an AI works alongside you in the terminal, reading code, executing shell commands, and choosing the next steps on its own. Kimi Code CLI is a terminal-based coding agent that uses Moonshot AI's Kimi model by default but allows you to configure other compatible providers. It deploys with a single command without requiring a Node.js installation, enabling you to analyze projects and perform tasks immediately without complex environment setup.
Single binary distribution and quick start
The tool is distributed as a single binary, allowing installation with one command without Node.js environment configuration or path issues. The terminal user interface is ready in milliseconds, so you do not feel any heaviness when starting a session. Afterward, you can enter the kimi command to open the interactive UI and perform initial tasks, such as requesting an explanation of the project structure.
Video input and interactive MCP configuration
If you add a screen recording or demo clip to the chat, the agent watches it to understand screen behaviors that are difficult to explain. It performs various tasks, such as converting reference videos to LUTs, editing long videos into shorter clips, and turning screen recordings into working code. You can add, edit, and authenticate Model Context Protocol servers interactively via the /mcp-config command without directly modifying JSON files. This simplifies the complex configuration process and increases agent extensibility.
Sub-agents and editor integration
You can run built-in sub-agents like coder, explore, and plan in parallel within isolated contexts while keeping the main conversation flow clean. This separates intensive tasks to improve processing efficiency. In editors that support the Agent Client Protocol, such as Zed and JetBrains, you can directly control sessions through the kimi acp subcommand. By adding this command to your editor settings, you can continue working on existing sessions without separate logins.
Execution environment and license
Windows users must install Git for Windows before the first run, and if Git Bash is in a custom location, you must set the KIMI_SHELL_PATH environment variable. To clone and run the source for development purposes, you need Node.js 24.15.0 or higher and pnpm 10.33.0. This project is released under the MIT license, allowing free use and modification. It is suitable for developers who prefer a terminal-centric workflow and want to integrate AI agents into their coding tasks.