Kimi Code CLI
A tiernew this weekMoonshot AI's open-source terminal coding agent that edits code, runs commands, and dispatches parallel subagents — Claude Code's scrappier, cheaper Chinese rival.
Kai's verdict
Solid open-source entry into the terminal agent race with genuinely interesting ideas — conversational MCP setup and parallel subagents are real differentiators — but it's brand new and the Kimi model ecosystem is less proven globally than Anthropic's or OpenAI's. Worth watching closely. (Verdict pending Phi's full review.)
Strengths
- Parallel subagents (coder, explore, plan) run in isolated contexts for faster multi-file work
- AI-native MCP configuration via conversational /mcp-config — no raw JSON hand-editing
- Multi-provider support: works with Moonshot, OpenAI, Claude, Gemini, or any OpenAI-compatible API
- ACP integration lets Zed, JetBrains, and VS Code drive CLI sessions directly from the IDE
- Single-command install with no Node.js pre-requisite, plus video/image input support in the TUI
Weaknesses
- Servers are primarily based in Asia, so latency can be noticeably higher for Western users
- Model quality still unproven relative to Claude Code / Codex at scale; early days for the new TypeScript rewrite
- Free tier has daily usage limits that can feel restrictive for heavy sessions
Best for
Terminal-native developers who want a Claude Code-style agentic workflow but with cheaper model costs, open-source flexibility, and the ability to swap in their own backend.
Pricing
Free (MIT open source) + Kimi Code membership for model access (~$19/mo)
CLI is free and MIT-licensed; model usage requires Kimi Code OAuth login or a Moonshot AI Open Platform API key. Can also be pointed at third-party providers (OpenAI, Claude, Gemini, OpenRouter).