Qwen Code

Qwen Code is an open-source AI coding agent that runs in your terminal. It's built for developers who want an agentic coding assistant without being locked into a single model provider, forked from Google's Gemini CLI and now developed independently with deep support for Qwen models alongside other providers.

Key features

• Agentic workflows – Auto-Memory, Auto-Skills, SubAgents, and Agent Teams work out of the box, with no manual configuration needed to get dynamic multi-step behavior.
• Multi-protocol support – Talks to OpenAI, Anthropic, Gemini, and Qwen APIs, plus any third-party provider or local model through Ollama or vLLM. You can switch providers at runtime.
• Open-source stack – Both the agent framework and the Qwen models behind it are open-source, so there's no vendor lock-in on either layer.
• Multiple interfaces – Beyond the terminal UI, it ships IDE plugins for VS Code, JetBrains, and Zed, a desktop app for macOS/Windows/Linux, a daemon mode (qwen serve) for shared sessions, and SDKs for TypeScript, Python, and Java.
• IM bot integration – Connects to Telegram, DingTalk, WeChat, and Feishu via qwen channel, letting you run coding tasks from chat apps.
• Headless mode – Run qwen -p "..." for scripting, CI/CD pipelines, and batch processing without the interactive UI.
• MCP and skill support – Includes MCP integration, Plan Mode, LSP integration, sandbox execution, and Git worktree support, plus built-in skills like /review, /batch, /loop, and /bugfix.
• Feature parity with Claude Code – The project tracks and aims to match Claude Code's capabilities (SubAgents, Hooks, Auto Mode, computer use) while adding multi-protocol and self-hosting options Claude Code doesn't offer.

Ideal use cases

Qwen Code fits developers who want a terminal-first coding agent they can point at different model backends depending on cost, latency, or privacy needs. It's a good match for teams already using local models (Ollama, vLLM) who want an agent that isn't tied to one vendor's API. If you need to script AI-assisted coding tasks into CI/CD, headless mode and the SDKs (TypeScript, Python, Java) make that straightforward.

It also suits developers who want to interact with a coding agent from IDEs (VS Code, JetBrains, Zed), a desktop app, or even messaging apps like Telegram or Feishu, rather than being stuck to a single terminal session. The daemon mode (qwen serve) is useful if multiple clients need to share one agent session over HTTP.

This is not a good fit if you want a fully managed, zero-maintenance SaaS product. You need Node.js 22+ installed, an API key or local model setup, and some comfort with terminal tools. If you specifically need a closed, single-vendor product with dedicated enterprise support, this open-source project may require more self-management than that would.

Installation

On Linux or macOS, install with the provided script:

curl -fsSL https://qwen-code-assets.oss-cn-hangzhou.aliyuncs.com/installation/install-qwen-standalone.sh | bash

On Windows, use PowerShell:

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

Restart your terminal afterward so environment variables take effect.

Alternatively, install via NPM (requires Node.js 22+):

npm install -g @qwen-code/qwen-code@latest

Or via Homebrew on macOS/Linux:

brew install qwen-code

Once installed, launch the interactive terminal UI and configure your provider:

qwen          # Launch interactive terminal UI
# Inside the session:
/auth         # Configure your provider and API key

For scripting or CI/CD, use headless mode instead: qwen -p "your prompt". Check the Authentication Guide and Settings Reference on the project's docs site for provider-specific setup details.