# Kodwai

> A platform where developers test and score their vibe coding skills by solving real challenges on their own machine using AI agents like Claude Code, Cursor, or Codex.

Kodwai is a coding challenge platform built specifically for the AI-agent era of software development. Rather than testing memorized algorithms, it scores how well a developer directs an AI agent through a real, ticket-sized problem — capturing the judgment, verification, and steering that define modern engineering work. The platform is fully free for developers and launched with 15 live challenges across 10 categories.

## What It Is

Kodwai positions itself as the first platform built to score "vibe coding" — the practice of building software by directing an AI agent rather than writing every line manually. Developers pick a challenge, run a CLI tool on their own machine, solve the problem using their preferred agent (Claude Code, Cursor, Codex, Aider, Cline, or any terminal-based agent), and submit for scoring. The platform then analyzes the full session — prompts, git history, test runs, and agent transcript — and produces a score across three axes: Direction, Outcome, and Lift.

## How the Scoring System Works

The scoring model is the core differentiator. A single careless one-shot prompt can make tests pass, but Kodwai's score is dominated by Direction (50 points), which measures how the developer steers, verifies, and decomposes the problem. Outcome (35 points) evaluates what actually shipped and whether it holds up. Lift (15 points) captures edge cases a one-shot prompt typically misses. Every signal cites its own evidence from the session transcript, so developers can see exactly why each axis scored the way it did.

## The Local-First Workflow

The workflow is deliberately local and frictionless:
- Run `npx @kodwai/cli challenge <slug>` to download the problem, starter files, tests, and initialize a git repo with the timer running
- Solve on your own machine in your own editor with your own agent — no browser sandbox or artificial constraints
- Run `npx @kodwai/cli submit` to package code, git history, test runs, agent transcript, and elapsed time for scoring

This bring-your-own-agent model means the environment matches how developers actually work day-to-day.

## Leaderboard, Badges, and Public Profiles

Every scored run contributes to a global, difficulty-weighted leaderboard. Developers earn a public profile at `kodwai.com/developers/[username]` showing their score, rank, earned badges, and preferred agents. Badges are awarded automatically for milestones (First Blood, Five Down), streaks (On Fire, Week Warrior, Monthly Machine), performance (Perfectionist, Speed Demon), and agent specialization (Claude Master, Cursor Pro). The profile is explicitly designed to be shareable with hiring managers as an alternative to take-home assignments.

## Audience and Hiring Track

The platform targets developers who want to demonstrate AI-era engineering skills, particularly those aiming for roles at large tech companies. The FAQ notes that solving challenges, scores, profiles, and the leaderboard are free for developers, while a hiring track for teams running interviews is the paid product — though no pricing details for that track are published on the site.

## Current Status

Kodwai launched with a Product Hunt listing (post ID 1115189) and is live at app.kodwai.com. The platform currently offers 15 challenges spanning backend, platform, data, AI, distributed systems, frontend, mobile, realtime, search, and security categories at easy, medium, and hard difficulty levels. The site footer references © 2026, indicating active development into that year.

## Features
- Score vibe coding sessions across Direction, Outcome, and Lift axes
- Local-first CLI workflow — solve on your own machine with your own agent
- Supports Claude Code, Cursor, Codex, Aider, Cline, and any terminal agent
- Real ticket-sized challenges across 10 categories and 3 difficulty levels
- Full session analysis: prompts, git history, test runs, agent transcript
- Per-signal evidence citations explaining each score component
- Global difficulty-weighted leaderboard
- Public shareable developer profile with score, rank, and badges
- Automatic badge awards for milestones, streaks, and agent specialization
- Hiring track for teams running AI-era interviews

## Integrations
Claude Code, Cursor, Codex CLI, Aider, Cline, Git, npm/npx

## Platforms
WEB, API, CLI

## Pricing
Freemium — Free tier available with paid upgrades

## Links
- Website: https://kodwai.com
- EveryDev.ai: https://www.everydev.ai/tools/kodwai