MoonshotAI/kimi-code

What it solves

Kimi Code CLI provides an AI-powered coding assistant that lives directly in the terminal. It eliminates the friction of switching between a browser and an editor by allowing an agent to read, edit, and execute code, run shell commands, and search files within the local development environment.

How it works

The tool operates as a terminal-based agent that integrates with Moonshot AI's Kimi models (and other compatible providers). It uses a Terminal User Interface (TUI) for interaction and can be extended via the Model Context Protocol (MCP) to add new skills and data sources. It also supports the Agent Client Protocol (ACP), allowing it to be driven by external IDEs like Zed or JetBrains.

Who it’s for

Developers who want a high-performance, terminal-centric AI coding agent that can perform complex tasks autonomously and integrate seamlessly with their existing IDE and shell workflow.

Highlights

• Autonomous Capabilities: Can read/edit code, run shell commands, and fetch web pages.
• Multimodal Input: Supports video input, allowing users to drop screen recordings or demo clips into the chat for the agent to analyze.
• Extensible Architecture: Features a rich plugin ecosystem and conversational MCP configuration.
• Parallel Workflows: Uses specialized subagents (coder, explore, and plan) to handle focused tasks in isolated contexts.
• IDE Integration: Connects to editors via ACP for a unified experience across the CLI and IDE.