Qwen Code

Qwen Code is an open-source coding agent from the Qwen team at Alibaba. It is an interactive CLI at its core, with headless mode, IDE integrations for VS Code, Zed, and JetBrains, and a desktop app built on top. It natively supports the OpenAI, Anthropic, and Gemini protocols, so it can connect directly to the ScitiX OpenAI-compatible endpoint. The configuration below applies to all Qwen Code forms. Qwen Code requires Node.js 22 or later for the npm install. If the install produces a broken version or Qwen Code will not start, check node --version; installing with Node below 22 can pull an incomplete build.

Configure

1. Install Qwen Code

Check your Node version first:

node -v

If it is below 22, upgrade Node. The official download page covers every platform and method.

brew install node@24

winget install OpenJS.NodeJS.LTS

nvm install 24 && nvm use 24
nvm alias default 24

With nvm, set a default alias. Without it, new terminals may fall back to the old Node version and npm-installed commands can disappear from PATH.

Install Qwen Code:

npm install -g @qwen-code/qwen-code@latest
# or
brew install qwen-code
qwen --version

2. Configure with environment variables

Create an API Key on the API Keys page, then put the connection settings in Qwen Code’s env file. It is loaded automatically on every launch and in every terminal.

mkdir -p ~/.qwen
cat >> ~/.qwen/.env <<'EOF'
OPENAI_API_KEY=<Your API Key>
OPENAI_BASE_URL=https://api.scitix.ai/model-api/v1
OPENAI_MODEL=glm-5.2
EOF
qwen

Plain export commands also work, but only last for the current shell session. Qwen Code searches .qwen/.env, then .env walking up from the current directory, then ~/.qwen/.env, then ~/.env. A project-level .qwen/.env lets different projects use different keys. On first launch, pick your provider in the /auth dialog. The current UI offers Alibaba ModelStudio, Third-party Providers, and Custom Provider. Use Custom Provider for ScitiX.

3. Configure with `settings.json`

For a persistent setup, create the config directory if this is a fresh machine, then edit ~/.qwen/settings.json:

mkdir -p ~/.qwen

{
  "modelProviders": {
    "openai": {
      "protocol": "openai",
      "models": [
        {
          "id": "glm-5.2",
          "baseUrl": "https://api.scitix.ai/model-api/v1",
          "envKey": "SCITIX_API_KEY"
        }
      ]
    }
  },
  "security": { "auth": { "selectedType": "openai" } },
  "model": { "name": "glm-5.2" }
}

Store the key in Qwen Code’s env file:

echo 'SCITIX_API_KEY=<Your API Key>' >> ~/.qwen/.env
qwen

A plain export SCITIX_API_KEY="..." also works, but only lasts for the current shell session. A new terminal loses it and Qwen Code shows “Missing credentials … Set that environment variable”. With security.auth.selectedType set, Qwen Code starts directly into the configured provider without the /auth dialog.

Field	Description
`modelProviders.openai.models[].id`	Model ID, such as `glm-5.2`. Must match `model.name` exactly.
`modelProviders.openai.models[].baseUrl`	ScitiX OpenAI-compatible endpoint: `https://api.scitix.ai/model-api/v1`.
`modelProviders.openai.models[].envKey`	Environment variable holding your API Key. Optional; defaults to `OPENAI_API_KEY`.
`security.auth.selectedType`	Set to `openai` to use this provider by default.

model.name must match an id under modelProviders exactly. Otherwise, the model will not show up in /model.

glm-5.2 is the recommended default. For faster responses, DeepSeek-V4-Flash is also verified to support native tool calling. If you do not want to export anything, settings.json also accepts a top-level "env": { "SCITIX_API_KEY": "..." } block. It has the lowest priority; shell exports and .env files win. A project-level .qwen/settings.json overrides the home-directory one.

Verify

In interactive mode, send a task such as “list the current directory and summarize the project structure” and confirm that tool calls and responses work. You can also verify headlessly:

qwen -p "introduce yourself in one sentence, and name the model you're connected to"

# headless run that's allowed to edit files
# approval modes: plan / default / auto-edit / auto / yolo
qwen -p --approval-mode yolo "create a file named check.txt containing ok"

VS Code Companion

Qwen Code has an official VS Code extension, Qwen Code Companion (qwenlm.qwen-code-vscode-ide-companion). Run /ide install inside Qwen Code to set it up, or install it from the marketplace and run /ide enable. It feeds workspace context, such as recent files, cursor position, and selection, to the agent and shows native diff review. Zed and JetBrains integrations are also available. See the official IDE integration docs.

Get Started

Models

Usage

API Guides

Tool Integrations

Configure

1. Install Qwen Code

2. Configure with environment variables

3. Configure with `settings.json`

Verify

VS Code Companion

​Configure

​1. Install Qwen Code

​2. Configure with environment variables

​3. Configure with settings.json

​Verify

​VS Code Companion

Configure

1. Install Qwen Code

2. Configure with environment variables

3. Configure with `settings.json`

Verify

VS Code Companion