Kimi CLI Cheat Sheet — frontier-class power, different economics

When Moonshot AI released Kimi K2 — a massive open-weight Mixture-of-Experts model — priced far below the frontier labs, the whole market looked up. Their CLI agent completes the picture: the same terminal-agent concept, with completely different economics.

Install & setup

Command	What it does
uv tool install --python 3.13 kimi-cli	Install (via Python's uv)
kimi	Open the interactive session
/login	Sign in with your Moonshot account (first run)
kimi --help	All flags

Inside the session

Command	What it does
/help	All commands
/model	Switch model
/clear	Reset conversation
/mcp	Manage MCP servers
Ctrl+X	Toggle direct shell mode — run plain terminal commands, then switch back
/exit	Quit

Why try it at all?

The pivotal point: cost. Kimi model API pricing sits well below Western frontier models, with coding-benchmark performance near the top tier. For high-volume daily work — refactoring, boilerplate, scripts — the end-of-month difference is very real.

And because the underlying Kimi models are open-weight, you can run them via other providers or even self-host with the right hardware — freedom most competitors don't offer.

The verdict

Run this experiment: take a real task from your job, run it through Kimi CLI and one other agent from these cheat sheets, compare output and the bill. Often "the best model in the world" is not "the best economic decision for this task." That mindset is exactly what our Models comparison page was built for.