samarailly51-pixel/opencode-harness

184 stars · Last commit 2026-06-14

Clean-room, model-agnostic harness for Claude Code-class coding agents

README preview

# OpenCode Harness

[![CI](https://github.com/samarailly51-pixel/opencode-harness/actions/workflows/ci.yml/badge.svg)](https://github.com/samarailly51-pixel/opencode-harness/actions/workflows/ci.yml)
[![Pages](https://github.com/samarailly51-pixel/opencode-harness/actions/workflows/pages.yml/badge.svg)](https://github.com/samarailly51-pixel/opencode-harness/actions/workflows/pages.yml)
[![Release](https://img.shields.io/github/v/release/samarailly51-pixel/opencode-harness?display_name=tag)](https://github.com/samarailly51-pixel/opencode-harness/releases/tag/v0.1.0)
[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](LICENSE)
[![Python](https://img.shields.io/badge/python-3.11%2B-blue.svg)](pyproject.toml)

OpenCode Harness is a clean-room, model-agnostic AI coding agent harness that standardizes how Claude Code / Codex-class coding agents execute tasks, call tools, produce traces, and generate evaluation reports.

中文定位:OpenCode Harness 是一个面向 AI coding agent 的开源运行与评测框架,用统一的 agent loop、工具权限、trace、eval suite 和报告系统,把一次性的 coding demo 变成可复现、可追踪、可对比、可诊断的工程流程。

This project does not contain or derive from Claude Code source code. It is an independent clean-room implementation of a coding-agent harness.

## Links

- Website: <https://samarailly51-pixel.github.io/opencode-harness/>
- DeepSeek case study: <https://samarailly51-pixel.github.io/opencode-harness/deepseek-case-study.html>
- Release: <https://github.com/samarailly51-pixel/opencode-harness/releases/tag/v0.1.0>
- 中文介绍: [docs/zh-intro.md](docs/zh-intro.md)

View full repository on GitHub →