Kimi Code CLI brings video-to-code conversion to open-source terminal agents
Moonshot AI released Kimi Code CLI, an MIT-licensed terminal agent that reads and edits code, executes shell commands, analyzes screen recordings, and integrates with major IDEs.
Moonshot AI released Kimi Code CLI this week, an open-source terminal agent for developers licensed under MIT. The tool reads and writes code, executes shell commands, searches for information, and plans tasks autonomously. It supports Model Context Protocol (MCP) for context management and Agent Client Protocol (ACP) for IDE integration, positioning it as an open alternative to Claude Code.
The standout feature is video input—developers can record screen interactions and have the tool convert them to code. Kimi Code CLI runs multiple sub-agents for coding, research, and planning, includes lifecycle hooks for auditing decisions and triggering automation, and offers a shell mode toggle (Ctrl-X) to switch contexts without exiting. The text-based interface loads in milliseconds and installs with a single command. It integrates with Zed, JetBrains IDEs (IntelliJ IDEA, PyCharm, WebStorm), and VS Code.
The default model is Kimi K2.6, a mixture-of-experts architecture with 1 trillion parameters that handles text, images, video, and PDF input. The MIT license aligns it with projects like Continue and Aider, and the sub-agent architecture lets teams parallelize research, code generation, and planning. Installation and integration guides are available on GitHub now.
