Kimi Code

Kimi Code is Moonshot AI’s coding-focused agent for terminal and IDE workflows. Its open-source Kimi Code CLI runs as the kimi command and can read and edit code, execute shell commands, search code, fetch web pages, and adjust its plan while working. It is installed with Moonshot’s script or as the kimi-cli Python package via uv, and it supports interactive terminal use, a local browser UI through kimi web, and agent-client integrations through kimi acp.

Top features

• Terminal coding agent: use natural-language prompts in a project directory to inspect files, modify code, run commands, and automate development tasks.
• Shell command mode: switch into shell-command mode with Ctrl-X to run commands without leaving the Kimi Code CLI session.
• IDE integration: connect through the Kimi Code VS Code extension or run kimi acp as an Agent Client Protocol server for ACP-compatible editors and IDEs.
• Browser UI: start kimi web for a local graphical interface with session management, file references, code highlighting, and related workflow features.
• MCP support: manage Model Context Protocol servers with kimi mcp, including stdio and streamable HTTP servers, OAuth authorization, and ad-hoc MCP config files.
• Project context: use /init to generate an AGENTS.md file for repository conventions, and use slash commands such as /login and /help during interactive sessions.
• Cross-platform install path: official docs cover Linux/macOS shell installation, Windows PowerShell installation, and uv tool install --python 3.13 kimi-cli; Python 3.12-3.14 is supported.

Use cases

• Implement features, fix bugs, and refactor code from the terminal.
• Explore unfamiliar repositories and ask architecture or implementation questions.
• Run builds, tests, scripts, and file-processing tasks as part of a guided coding session.
• Bring a terminal coding agent into VS Code, Zed, JetBrains, or another ACP-capable editor.
• Connect external tools through MCP for workflows that need additional local or remote capabilities.