Qwen launches Qwen Code 1.0 Preview as an open-source agent harness
Qwen Code 1.0 Preview combines Qwen Code and Qoder capabilities in an open-source terminal harness for coding and automated workflows.
Qwen has released Qwen Code 1.0 Preview as an open-source agent harness for software development, bringing together capabilities from Qwen Code and Qoder in a terminal-focused tool. The preview arrived on 11 November 2026 and is positioned for both interactive coding sessions and automated workflows.1
What Qwen Code 1.0 Preview does
The release notes say the tool can help with code understanding, implementation, debugging and testing. It supports Qwen and Qoder accounts, multiple model providers and compatible custom endpoints, giving developers more choice over which model service sits behind the agent.1
Qwen also says the preview ships as native packages for macOS, Linux and Windows. That means users do not need to install Node.js separately before trying the tool, according to the release announcement. The package downloads are provided through the project's GitHub release page.1
The preview is designed to preserve work between sessions. Qwen lists session resumption and non-interactive execution among its headline features, which should make the tool relevant to scripted development tasks as well as a developer working directly in a terminal.1
An agent harness rather than just a coding chatbot
Qwen's documentation describes Qwen Code as a terminal-based agentic coding tool and calls the 1.0 Preview an open-source harness for complex tasks.2 The product site presents a wider execution model built around terminal and automated workflows, tool execution, permission controls, MCP connections, multi-agent teams and reusable workflows.2
That distinction matters. The announcement is not simply for another assistant that suggests code in a chat window. Qwen is presenting a runtime that can manage context, call tools and continue a task across a longer workflow. The documentation also describes a persistent AppServer model in which sessions keep their own execution state while model interfaces and tool connections remain reusable.2
The preview status is important, however. Qwen warns that interfaces, configuration options and behaviours may change. Developers considering it for production automation should therefore treat the current release as an evaluation build and test its permission boundaries, model-provider configuration and session handling before putting it into unattended jobs.1
Why the Qoder connection matters
Qwen says Qwen Code 1.0 Preview brings together the agent capabilities of Qoder and Qwen Code. The official documentation describes the result as a reusable agent runtime that can work with models, tools and execution environments, while the release notes focus on the practical coding tasks it can perform.12
For developers, the most notable promise is the combination of a local terminal interface with automation-friendly execution. A tool that can resume sessions and run non-interactively may fit code maintenance, testing and repeatable repository tasks more naturally than a purely conversational product. Whether it delivers that promise consistently will depend on the preview's stability and the quality of the configured model and tools.
Our opinion
Qwen Code 1.0 Preview is an interesting move because it treats AI-assisted programming as an ongoing execution environment, not merely a prompt box. Native packages, session resumption and non-interactive execution give it a credible practical angle from day one. The sensible approach is to experiment with it on contained repositories and review its permissions carefully, rather than assuming a preview runtime is ready to own critical production changes.