Kimi Code CLI Cheatsheet - Moonshot Terminal AI Coding Agent

Kimi Code CLI is Moonshot AI terminal AI coding agent, wired out of the box to Kimi models, for developers in the Kimi ecosystem who want to edit code and ask questions with Kimi directly in the terminal. It is optimized for Kimi out of the box, removing manual model wiring. This sheet covers install, interactive and -p single instructions, session resume, and /model switching.

AI CLI·34 commands·Last updated 2026-08-26

Install & Launch 4

npm install -g @moonshot-ai/kimi-code

Install globally via npm

curl -fsSL https://code.kimi.com/kimi-code/install.sh | bash

Official script install (no Node.js prereq)

kimi --version

Print the installed version

kimi

Start the interactive UI in the project dir

Interactive & Single Instructions 5

kimi -p "查看此项目的目录结构"

Run a single instruction without the interactive UI

kimi -C

Resume the previous session

/login

/logout

Clear the current credentials

/init

Scan the project and generate AGENTS.md

Sessions & Models 5

/model

Switch the current model

/compact

Manually compact context to free tokens

/fork

Fork the current session, keep history, continue independently

/sessions

Browse session history and pick one to resume

kimi acp

Drive Zed/JetBrains editors via Agent Client Protocol

Subcommands & Ops 20

kimi --session <id> / -S

Resume a session by ID; picker without ID

kimi --model <model> / -m

Specify the model alias for this launch

kimi --yolo / -y

Auto-approve regular tool calls (trusted dirs only)

kimi --auto

Start in auto permission mode; agent asks nothing

kimi --plan

Start a new session in Plan mode

kimi --skills-dir <dir>

Load Skills from a directory (repeatable)

kimi --add-dir <dir>

Add an extra workspace dir (repeatable)

kimi --work-dir <dir> / -w

Set the root for file operations

kimi --thinking / --no-thinking

Explicitly enable/disable thinking mode

kimi --agent <name>

Select a built-in agent spec (experimental)

kimi login

Non-interactive device-code OAuth login

kimi server run

Start or reuse the local Kimi service daemon

kimi server status

Print install/run status and port (--json for scripts)

kimi web

Start service in background and open the web UI

kimi doctor

Validate config.toml and tui.toml

kimi doctor config

Validate only config.toml

kimi doctor tui

Validate only tui.toml

kimi export [sessionId]

Package a session as a ZIP for sharing

kimi upgrade

Check and upgrade to the latest version

kimi provider

Manage provider config

Tips

Kimi Code CLI ships as a zero-dependency binary: one script install and it just works — no PATH fiddling, no conflicts with global modules.
`kimi -p` runs a single instruction and `kimi -C` resumes a session — these are the workhorses for scripts and automation. Run `/init` before big tasks to generate AGENTS.md so the agent understands the project.
To use other vendors (Anthropic/OpenAI/Google), just edit `~/.kimi-code/config.toml` with the corresponding API key.

FAQ

What is the relationship between the kimi command and the kimi-cli package?

This sheet covers Moonshot official Kimi Code CLI (npm package @moonshot-ai/kimi-code); the installed command is `kimi`. There is also a separate community Python package named kimi-cli, which is a different project.

Can Kimi Code CLI run on Windows?

Yes. On Windows run the install script `irm https://code.kimi.com/kimi-code/install.ps1 | iex` in PowerShell. Install Git for Windows first; the CLI uses its Git Bash as the shell environment.

Official References

Each command links to its official documentation below, so you can verify the latest usage and read deeper.

Maintained by LaoHand

Publicly updated on Aug 26, 2026, continuously proofread against official docs.

Contact Us

Wrong command or description? Send us corrections, business inquiries or product feedback by email.