Alibaba has officially released Qwen Code 1.0, integrating the Harness capabilities previously developed for Qoder into an open-source framework. This version is designed to provide developers with a foundation for embedding AI agents directly into their own software products.
The architecture of Qwen Code 1.0 utilizes an in-memory event bus that supports evolution toward distributed systems. By separating reasoning orchestration from tool execution, the system transitions from a single-process, multi-session model to a multi-node, multi-session architecture.
The platform features a unified event mechanism that manages connections between tasks and system components. This approach allows for modular expansion of reasoning, memory, and tool capabilities while maintaining centralized control over model services, sandbox resources, and MCP configurations.
Developers can leverage the integrated Harness kernel to connect AI models with external tools and execution workflows. The Qwen Code 1.0 Preview is currently available as an independent branch, while the existing stable branch remains supported by the development team.