Image: QwenLM
Why it mattersA team picking a coding agent now has a fourth option that runs with no Node install. The agent features are open, so a buyer can read them.
QwenLM released Qwen Code 1.0 Preview on 11 November. It is a terminal coding agent that runs on your own machine. The project is open source under Apache 2.0. It has 29,931 stars on GitHub. Trendshift ranked it 17th on its daily list on 12 November.
The 1.0 Preview is the first release to pull in features from Qoder. Qoder is Alibaba's paid coding IDE. The release notes say the new version brings together the agent features of Qoder and Qwen Code. It covers code understanding, implementation, debugging and testing. That work was in the paid product. It is now in the open-source tree.
What the release notes list
QwenLM names four things in the release. The first is the Qoder merge above. The second is support for Qwen accounts, Qoder accounts, several model providers and custom endpoints in the OpenAI format. The third is native packages for macOS, Linux and Windows. These run without a separate Node.js install. The earlier version needed Node. The fourth is session resumption and non-interactive execution. A run can be scripted or picked up later.
The agent features the README lists include Auto-Memory, Auto-Skills, SubAgents and Agent Teams. Auto-Memory keeps context across turns. SubAgents and Agent Teams let the main agent hand off to smaller agents. The README says the framework also speaks OpenAI, Anthropic and Gemini APIs, not only Qwen.
A follow-up, v1.0.0-preview.1, was released the next day. It fixed Qwen subscription accounts. It added parallel tool calls for faster multi-tool work.
How this compares to the alternatives
A developer picking a terminal agent today chooses from Claude Code, OpenAI Codex and a few open projects. Claude Code and Codex are the two most used. Both are closed source. Qwen Code is the one in that group whose harness is in a public repository. A team can read what it does. They do not have to trust a release note.
The 29,931 star count puts it behind Claude Code and Codex in the public numbers. It is ahead of most other open terminal agents. Trendshift ranked it 17th on 12 November.
What it costs to try
The install step changed in 1.0. The native binary route needs no Node and no npm install. The source route still exists. Qwen Code works with a free Qwen account or a paid one. It can also point at an OpenAI-shaped endpoint. That lets a team run it against a model they already pay for. The release page lists ten downloads for the main desktop operating systems.
QwenLM calls 1.0 a Preview. The release notes warn that interfaces, configuration options and behaviour may change during Preview. A team picking it for daily work should track the release page for a few more weeks. The behaviour will settle.
A team already using Claude Code or Codex now has a fourth agent to compare on its own work. This one lets them read the code. A team that built around Qoder gets the same agent features without the IDE and without the subscription.
Source
- Qwen Code v1.0.0-preview release notes, QwenLM
- QwenLM/qwen-code repository, QwenLM
This item was written by an AI system from the linked source. Reveneau is responsible for what it publishes.
Get AI News in your inbox
New developer tools, model and agent releases, and how teams are actually using them to release software. Short, and only when there is something worth reading.
