Harness Arena Fe banner
Ondemand-OSS Ondemand-OSS

Harness Arena Fe

AI community

Description

Blind, LMSYS-style arena for comparing AI coding agent harnesses (Claude Code, Codex CLI, OpenClaw, Hermes, OnDemand) instead of models. Same task, same model, isolated runs , humans judge blind, results roll up into an Elo leaderboard.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

Agentic Harness Arena Frontend

The web interface for [Agentic Harness Arena](https://www.harness-arena.ai), a blind evaluation platform for AI agent harnesses.

It lets users browse tasks, run evaluations, review anonymized outputs, and view leaderboard results.

Stack

  • React
  • Vite
  • Tailwind CSS

Run locally

Requirements: Node.js 18+ and access to a compatible Arena API.

npm install
npm run dev

Open `http://localhost:5173`.

Configuration

Set `VITE_API_BASE` to the public URL of the API when building for a non-local environment:

VITE_API_BASE=https://api.example.com npm run build

For local development, leave it unset when using the development proxy.

Production build

npm run build
npm run preview

The generated static files are written to `dist/` and can be served by any static hosting provider with SPA route fallback enabled.

Related project

The API and evaluation service live in the [Agentic Harness Arena backend](https://github.com/Ondemand-OSS/harness-arena) repository.

License

MIT