Kimi CLI

Kimi CLI is the command-line agent product RC discusses in 用 Agent 动力学，和 40 个 Agents 一起为「人 + AI」做产品｜对谈 Slock.ai 创始人 RC. RC describes it as a coding/general agent running in the command line and distinguishes it from the other CLI tools that such an agent may call.

The source frames Kimi CLI as both a product and a design probe. RC says he wanted the CLI form to be only the first layer: underneath it, a local Agent Harness could be reused and then wrapped into SDK, Web UI, or VS Code extension forms. That makes Kimi CLI relevant to the wiki less as a standalone feature list and more as an example of how Agent-Facing Interfaces and Agent-Optimized CLI design can expose model capability to real tools.

Design Lessons

• CLI is useful for agents because language models handle text streams, command output, and terminal conventions well.
• Agent-facing commands need compact inputs, stable and information-dense outputs, clear documentation, and examples the model can reuse.
• A coding agent can begin from a small agent loop and a Bash tool, then gain reliability as missing tools and prompt rules are discovered from failures.
• CLI is not necessarily the final consumer interface; it can be the first accessible harness surface before more human-friendly UI layers appear.

Connections

• Kimi — product/company context for Kimi CLI.
• RC — builder and source speaker.
• Agent-Optimized CLI, Agent-Facing Interfaces, and Agent Harness — core interface and execution concepts the source reinforces.
• Vibe Coding, Claude Code, Codex, and Model Workflow Fit — adjacent coding-agent workflow comparisons.