Overview
Qwen Code is an open-source coding agent from the Qwen team at Alibaba. It runs in your terminal like Claude Code or Gemini CLI, reads your repository, and makes changes on request. It isn't tied to Qwen models: it speaks the OpenAI, Anthropic, Gemini and Qwen APIs and can run local models. A 1.0 preview merges in agent-harness features from Qoder.
Key capabilities
- Agent features: auto-memory, auto-skills, subagents, agent teams and MCP support.
- Many surfaces: terminal, IDE plugins, desktop app, web UI and chat apps.
- Headless mode:
qwen -p "..."for scripts and CI jobs. - SDKs: TypeScript, Python and Java for building on top of it.
- Daemon mode:
qwen serveshares sessions across clients over HTTP and SSE (experimental).
Tradeoffs
- Fast-moving project; some interfaces (web UI, daemon) are marked experimental.
- The daemon listens without authentication by default; set a token on shared machines.
- Quality depends on the model you connect.
Ideal for
- Developers who want an open-source alternative to Claude Code
- Teams running local or non-OpenAI models
- CI automation with a coding agent
Not ideal for
- People who want a fully managed IDE experience
- Shared hosts without setting up auth
Getting started
- Install with
npm install -g @qwen-code/qwen-code@latest(Node.js 22+),brew install qwen-code, or the standalone install script. - Open a terminal in your project and run
qwen. - Run
/authto sign in with a Qwen subscription or configure an API key for OpenAI, Anthropic, Gemini or a local model. - Start with a prompt such as "Explain this repository and show me where to start." For scripts and CI, use headless mode:
qwen -p "...".
Technical notes
- Install:
npm install -g @qwen-code/qwen-code@latest(Node.js 22+) orbrew install qwen-code - Run:
qwenin your project - License: Apache-2.0
- Source: QwenLM/qwen-code
- GitHub stars: 29,505 (as of 2026-11-12)