How to Use Qwen Code: Installation, /auth, Japanese Localization, and API Settings

Qwen Code is an open-source AI coding agent (Apache-2.0) published by Alibaba's Qwen team. The workflow has four stages: “install → start qwen in a project → configure the connection with /auth → request a task.” However, because the free Qwen OAuth allocation ended on April 15, 2026, logging in as described in older guides will not make it run for free.This page explains, in order, how to get started using the current /auth menu, switch to Japanese, configure a custom connection for Claude or GPT, and use common commands. The latest version is v0.24.7, published to npm on September 29, 2026. Releases occur at least weekly, so check your local version with qwen --version.

Installation

These are the official README instructions. With the standalone installer, you do not need to provide Node.js yourself. Node.js 22 or later is required only when using npm.

# macOS / Linux（公式のスタンドアロン インストーラー）
curl -fsSL https://qwen-code-assets.oss-cn-hangzhou.aliyuncs.com/installation/install-qwen-standalone.sh | bash

# Windows(PowerShell)
irm https://qwen-code-assets.oss-cn-hangzhou.aliyuncs.com/installation/install-qwen-standalone.ps1 | iex

# npm（Node.js 22 以上が必要）
npm install -g @qwen-code/qwen-code@latest

# Homebrew(macOS / Linux)
brew install qwen-code

After installation, reopen the terminal so environment variables take effect. Qwen Code was originally based on Google Gemini CLI v0.8.2, but development diverged from upstream starting with v0.1. Gemini CLI settings and free allocations do not apply unchanged.

First launch and switching to Japanese

cd /path/to/your-project
qwen

# セッション内で：
/language ui ja-JP          # 画面表示を日本語に
/language output Japanese   # モデルの回答を日本語に
/auth                       # 接続先と API キーを設定

/language ui ja-JP changes the interface to Japanese, after which screens display Qwen Code's own Japanese labels such as “認証方法を選択” (Select an authentication method) and “ツール承認モード” (Tool approval mode). This page uses the same labels. The response language is separate from the UI language and is set with /language output Japanese.

In addition to the terminal, Qwen Code provides a desktop app, a browser-based Web UI (qwen serve --open, experimental), integrations for VS Code, Zed, and JetBrains, and headless execution for scripts and CI (qwen -p "..."). All are included in the same free repository; inference costs are separate.

Choose a connection with /auth

Opening /auth (also called /login) displays the “Select an authentication method” screen, with three top-level choices. If you try to select Qwen OAuth, it displays “Ended — switch to Coding Plan or API Key.”

The three ModelStudio options are not merely different ways to pay the same bill. The documentation assigns each a separate endpoint and separate key, and they will not work unless the key type matches baseUrl. As of October 1, 2026, Coding Plan is labeled “limited quantity, first come first served; replenished daily at midnight (UTC+8),” so it may be available on some days but not others. A single Coding Plan request internally becomes multiple model calls; Alibaba says simple tasks consume 5–10 calls, while complex tasks consume 10–30 or more. The English Qwen Code pricing page covers how to compare prices in detail.

Using Claude or GPT: configure Custom Provider in settings.json

Qwen Code's authentication documentation recommends ~/.qwen/settings.json's modelProviders for connecting to third parties such as OpenAI, Anthropic, Google, OpenRouter, and your own endpoint. For Kunavo, use the following:

{
  "modelProviders": {
    "openai": [
      {
        "id": "claude-sonnet-5",
        "name": "Claude Sonnet 5 (Kunavo)",
        "baseUrl": "https://api.kunavo.com/v1",
        "description": "Kunavo, OpenAI 互換",
        "envKey": "KUNAVO_API_KEY"
      }
    ]
  },
  "security": {
    "auth": {
      "selectedType": "openai"
    }
  },
  "model": {
    "name": "claude-sonnet-5"
  }
}

# キーは settings.json に直接書かず、.qwen/.env か環境変数に
echo 'KUNAVO_API_KEY=sk-kn-...' >> ~/.qwen/.env

There are four points to keep in mind.

• baseUrl only goes as far as /v1.The model provider reference explicitly says to specify the API's /v1 route, not /v1/chat/completions; the SDK appends the request path. Including the path produces a 404 rather than an authentication error.
• Where to put the key.Qwen Code reads from the environment variable named in envKey. The priority is the shell's export, then .env (.qwen/.env recommended; only the first file found), then settings.json's env, with the last option discouraged because it stores the key in plain text.
• The configuration file takes precedence over CLI flags.The resolution order is the value entered through /auth → the selected modelProviders → CLI arguments → environment variables → settings.json. This is why --openai-base-url may appear to be ignored. security.auth.apiKey and security.auth.baseUrl in older guides are deprecated.
• Start with Chat Completions.If you omit wireApi, the Chat Completions format is used. Setting "responses" disables both automatic endpoint detection and fallback. Edits to modelProviders take effect immediately in running sessions (they appear after reopening /model).

Kunavo's catalog does not include Qwen text models.This is not a way to use Qwen cheaply; it is a way to use Claude or GPT inside Qwen Code with one prepaid balance. Kunavo has also not actually tested Qwen Code against its own endpoint. As with the English Qwen Code configuration page, this configuration was created from the official documentation. Keep the route that currently works and test alongside it.

Your first task and common commands

As in the README example, start with a request such as “Explain this repository and tell me where to start reading.” This checks that file reading and tool calls work end to end. A simple greeting will not reveal connection problems.

Automatic approval modes such as yolo come with the documentation's warning to use them only in “trusted, sandboxed, or disposable environments.” The estimated cost shown by /stats model is calculated by Qwen Code and is not the billed amount. Check the connection's usage records for the actual amount.

Limitations to know

• The built-in web_search is determined by the connection.Because it uses DashScope's server-side search, it is enabled for ModelStudio Standard API Key and Token Plan connections, and for entries pointing to recognized DashScope hosts; it is disabled for Coding Plan (not verified for that endpoint) and for custom endpoints at third-party or other hosts (web_search documentation). Add an MCP search server if needed.
• Kunavo's route is chat-only.Kunavo has no embedding, text-to-speech, or speech-recognition models. Qwen Code Live Voice requires a DashScope endpoint, so regardless of where the chat model points, Live Voice still requires a separate key.

Payment when trying Kunavo

Kunavo uses a prepaid top-up system with no monthly fee, deducting charges from your balance by token. The minimum top-up is $10, and Stripe Checkout accepts cards (Visa, Mastercard, American Express, JCB), Apple Pay, Google Pay, and Link. Review the billing explanation, then create an account and generate a key. After your first request, the most reliable way to verify the actual amount charged is to check your usage records.

Frequently asked questions

Verified October 1, 2026: Qwen Code's README; the authentication, model provider, command, and web_search documentation (main branch); Japanese UI strings (packages/cli/src/i18n/locales/ja.js); npm @qwen-code/qwen-code 0.24.7; and Alibaba Cloud's Coding Plan and Token Plan pages. Kunavo has not run Qwen Code against its own endpoint.