Qwen Code cheat sheet

Every command, flag and slash command for qwen — Alibaba's terminal agent — forked from Gemini CLI but now its own tool, with ~60 slash commands. 17+ commands, verified July 2026.

Qwen Code hard-forked from Gemini CLI at its own v0.1 and diverged: the --all-files / -a flag is gone (use --include-directories), and the Qwen free tier was discontinued 2026-04-15 — use any OpenAI-compatible key or the Alibaba Cloud Coding Plan.

Install

npm install -g @qwen-code/qwen-code@latest   # npm (needs Node 22+)
brew install qwen-code                        # Homebrew
npx @qwen-code/qwen-code                       # run without installing

Start & run

qwen                              # interactive terminal UI
qwen -p "fix the failing tests"   # headless
qwen -m qwen3-coder-plus -p "…"   # pick the model
qwen --include-directories src,tests -p "…"   # add dirs to context
qwen sessions list                # list recent sessions

Key flags

Flag	What it does
`-m, --model <id>`	Model (e.g. `qwen3-coder-plus`)
`-p, --prompt <text>`	Headless one-shot prompt
`-i, --prompt-interactive`	Interactive with an initial prompt
`-y, --yolo`	Auto-approve all actions
`--approval-mode <m>`	`plan` · `default` · `auto-edit` · `auto` · `yolo`
`-s, --sandbox`	Sandbox (docker / podman / sandbox-exec)
`--include-directories <a,b>`	Extra dirs (the `--all-files` replacement)
`-o, --output-format <fmt>`	`text` · `json` · `stream-json`
`-c, --continue` / `-r, --resume`	Resume sessions

Slash · @file · !shell

Command	What it does
`/help` · `/clear` (`/new`)	Help · clear context
`/compress` (`/summarize`)	Summarize history to save tokens
`/stats` · `/context`	Usage · context-window breakdown
`/memory` · `/remember` · `/forget`	Memory manager
`/model [id]` · `/approval-mode`	Switch model · change approval mode
`/init`	Analyze the dir and create a `QWEN.md`
`/review [PR]`	Code review with parallel agents
`@file` · `!cmd`	Inject a file · run a shell command

Config

~/.qwen/settings.json   ·   ./.qwen/settings.json     # user / project
QWEN.md                          # project context (/init to scaffold)

export OPENAI_API_KEY="…"                             # any OpenAI-compatible provider
export OPENAI_BASE_URL="https://openrouter.ai/api/v1" # e.g. OpenRouter / Ollama / vLLM
export OPENAI_MODEL="qwen3-coder-plus"

Common one-liners

qwen -p "Write a concise commit message" --output-format json | jq -r '.result'
OPENAI_BASE_URL="http://localhost:11434/v1" OPENAI_API_KEY="ollama" \
  qwen -p "refactor utils.py"                          # local Ollama, one run
qwen -m qwen3-coder-plus --include-directories src,tests -p "fix the flaky test"
git diff origin/main...HEAD | qwen -p "Review these changes for bugs"