Coding agent (CLI) · Alibaba (Qwen)
Qwen Code
Terminal coding agent from the Qwen team that started as a fork of Gemini CLI v0.8.2 and has developed independently since v0.1. Works with several API protocols and local models.
Qwen Code review
Qwen Code is an open-source CLI coding agent from the Qwen team at Alibaba. The software is free under Apache-2.0 and runs on Windows, macOS and Linux. We read the vendor's docs and repository in October 2026. We did not run an independent test. The free hosted model it once offered is gone.
Who it is for
Developers who want an open-source terminal coding agent in the style of Claude Code. They should be ready to bring a provider key, a plan or a local model.
What stands out
Qwen Code began as a fork of Gemini CLI v0.8.2 and has developed on its own since v0.1. Install with npm, which needs Node.js 22 or newer, or with a standalone installer. Homebrew also has a formula. The npm package was at 0.26.2 on 27 October 2026. A desktop app, IDE plugins for VS Code, Zed and JetBrains, and an experimental web UI also exist.
It is not tied to Qwen models. It lists OpenAI-, Anthropic- and Gemini-compatible endpoints, DeepSeek, Kimi, MiniMax, Z.AI and OpenRouter. The README names Ollama and vLLM for local models.
Headless `qwen -p` runs prompts in scripts and CI. The README table lists subagents, MCP, sandbox and git worktrees. There is also an experimental `qwen serve` daemon, with chat channels for Telegram, DingTalk, WeChat and Feishu. SDKs exist for TypeScript, Python and Java.
Price and licence
The CLI is free under Apache License 2.0, plain text with no extra terms, so it is open source. Model cost comes from the provider you connect. Options are your own API key, Alibaba Cloud Coding Plan, Token Plan or a local model. The Coding Plan is a fixed monthly fee, and the docs page shows no price. We could not confirm tier prices.
Where it falls short
- The free Qwen OAuth tier ended on 15 April 2026, per the auth docs. Without it you need a paid key, a plan or a local model.
- Versions are fast-moving 0.x, with preview releases tagged almost daily.
- The npm route needs Node.js 22 or newer. The web UI and daemon mode are marked experimental.
- The docs do not describe running the agent on a remote machine over SSH.
- We checked sandbox details only in the README table, not on the dedicated docs page.
Alternatives
Its upstream is Gemini CLI. Compare it with OpenCode and the rest of the terminal coding agents.
Facts: platforms, price and licence
- platforms
- macOSWindowsLinuxStandalone installers for macOS/Linux and Windows; npm route needs Node.js 22+. Also ships a desktop app, IDE plugins (VS Code, Zed, JetBrains) and an experimental web UI.
- price
- FreeThe CLI itself is free (Apache-2.0). Model cost comes from whichever provider you connect: your own API key (OpenAI-, Anthropic-, Gemini-compatible, DeepSeek, OpenRouter and others), Alibaba Cloud Coding Plan (fixed monthly fee, price not on the docs page) / Token Plan, or local models. The earlier Qwen OAuth free tier was discontinued on 2026-04-15 per the auth docs (seen 2026-10-28).
- source
- Open sourceLicence: Apache-2.0
- github
- QwenLM/qwen-code 28.8k stars, seen 2026-10-28
Qwen Code features
- ✓ yes MCP
- ✓ yes Multi-agent
- ✓ yes Headless / CI
- ✓ yes Sandbox
- ✓ yes Local models
- ✓ yes Your own API key
- ✓ yes Git worktrees
- ? not verified SSH
✓ documented · – absent or not in its feature list · ? not verified
Qwen Code strengths
- Apache-2.0 licence, very active repo (pushed daily, 28k+ stars) with SDKs for TypeScript, Python and Java
- Not tied to Qwen models: OpenAI-, Anthropic-, Gemini-compatible endpoints, third-party providers and local servers such as Ollama/vLLM are listed
- Headless `qwen -p` for scripts and CI, plus an experimental `qwen serve` daemon and chat channels (Telegram, DingTalk, WeChat, Feishu)
- Documents subagents, agent teams, MCP, sandbox and git worktrees
Qwen Code weak points
- The free Qwen OAuth tier ended on 2026-04-15, so using it now means supplying a paid provider key, a Coding/Token Plan or a local model
- Fast-moving 0.x versions (0.26 at time of check) with preview releases tagged almost daily
- Needs Node.js 22+ when installed through npm; the web UI and daemon mode are marked experimental
- Docs do not document running the agent on a remote machine over SSH
Who Qwen Code is for
- best for
- Developers who want an open-source Claude Code-style agent that can point at Qwen, other cloud APIs or a local model.
- look elsewhere if
- Anyone looking for a zero-cost hosted model out of the box, now that the free OAuth quota is gone.
Compare Qwen Code
Lists that include Qwen Code
Sources and what we could not verify
Checked 2026-10-28. Prices and features change; the vendor's page is the authority.
- github.com/QwenLM/qwen-code — README: install, features table (sandbox, worktrees, headless, subagents, MCP), Gemini CLI v0.8.2 origin, local models Ollama/vLLM, platforms
- api.github.com/repos/QwenLM/qwen-code — stars, license Apache-2.0, pushed_at, archived=false
- github.com/QwenLM/qwen-code/blob/main/LICENSE — licence text (plain Apache 2.0, no extra terms)
- qwenlm.github.io/qwen-code-docs/en/users/configuration/auth — auth options, free OAuth discontinued 2026-04-15, Coding Plan / Token Plan, API keys
- qwenlm.github.io/qwen-code-docs/en/users/overview — overview, Node 22+, headless qwen -p, install
- registry.npmjs.org/@qwen-code%2Fqwen-code — latest npm version/date
Not verified
- Exact price of Alibaba Cloud Coding Plan / Token Plan not on the docs page
- Sandbox type and details only checked via the README table, not the dedicated sandbox docs page
- Whether local-model setup is walked through in the docs (README names Ollama/vLLM; auth docs only mention Custom Provider)
- ssh/remote agent execution not documented on the overview page