Qoder, Alibaba Cloud's intelligent agent coding platform, published a blog post on its official WeChat account yesterday (July 28), announcing the open-sourcing of Better Harness on GitHub.
According to the blog post, as cited by IT Home, Better Harness is an open-source analysis and continuous improvement tool for Coding Agent workflows. It connects the engineering practices, evaluation models, and runtime capabilities of Harness Engineering and Loop Engineering, and supports Claude Code, Codex, Qoder, and Cursor.
In terms of functionality, Better Harness does not consider the mere presence of a configuration as proof that a capability is working. It checks whether the Agent uses the relevant capability and whether that capability supports task completion. Each Finding must include traceable evidence, user impact, remediation scope, and a validation method.
In terms of open sourcing, the project adopts a three-layer open-source framework:
The first layer is Harness Engineering practices, covering Sessions, CLI, observability, Rules, Skills, MCP, Memory, Hooks, and automation.
The second layer is the Agent Work Loop evaluation model, which is responsible for turning engineering practices into questions that can be checked individually, while constraining the relationships among evidence, scores, and conclusions.
The third layer is an executable engineering implementation, ensuring that the first two layers do not remain limited to documents and models but can instead be run repeatedly in real projects.
The tool independently collects three types of evidence, including Session Evidence, Project Harness, and Agent Customize. The first round of internal evaluation covered 30 real GitHub projects.
Reference address above
Better Harness's GitHub repository