# Paweł Huryn releases Mini Harness, a 1,200-line open-source coding harness

> Original title: Everyone is discussing what a harness is. Nobody shows you how to build one. And everyone can.

- Source: X search: indie AI launches (I built, solo founder)
- Published: 2026-11-05T11:57:10.000Z
- Bad Influence: https://badinfluence.news/items/hs13i4yo64x3dzykmajdyh0nj
- Original: https://x.com/PawelHuryn/status/2118311658867097991

## Summary

Paweł Huryn introduces Mini Harness, an open-source coding harness of about 1,200 lines of Python that he built as a Claude Code alternative. He says it includes /goal mode, permissions with auto mode, skills, MCP servers, and Anthropic's server-side compaction. In a Terminal-Bench 2.0 chart he shares, Mini Harness passed 9.2 of 10 hidden tests with Claude Haiku 5.5 at high effort, ahead of seven other harnesses.

## Why it matters

The post names the open-source harness's features and the Terminal-Bench 2.0 result, which helps readers compare agent harnesses on Claude Haiku 5.5.

## Post

Everyone is discussing what a harness is. Nobody shows you how to build one. And everyone can.

So, I built my own Claude Code: https://github.com/phuryn/mini-harness

It's called Mini Harness: about 1,200 lines of Python.

Other minimal harnesses exist, but none has /goal mode, permissions + auto mode, skills and MCP servers. I wanted a fully working one to explain how a harness is built, down to Anthropic's server-side compaction. 

Surprisingly, it became the top harness on Claude Haiku.

How it works: 🧵

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWW8WslWQAAmfsc.jpg&mode=thumb&exp=1794096000&sig=e0a5f300067b9ec7)

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWW8ZCrWEAAxCD1.jpg&mode=thumb&exp=1794096000&sig=5214d26c3f06d8f9)