What is Kodwai?
Kodwai is the first platform that scores how well you collaborate with AI coding agents like Claude Code, Cursor, and Codex. Instead of testing what you memorized, it measures your ability to direct an agent through real coding challenges. You solve problems in your own terminal, and the CLI captures your code, tests, git history, agent transcript, and time. The platform then scores you across three axes—Direction, Outcome, and Lift—each backed by specific evidence. It's free, and you bring your own agent.
Who it's for
- Vibe coders who want a tangible metric for how effectively they steer AI agents through real problems
- Hiring managers tired of LeetCode scores that reveal nothing about how an engineer works with modern tools
- Developers building their profile who need proof of their engineering judgment, not just memorized syntax
Key features
Real challenges on your machine
Kodwai provides ticket-sized problems across categories you actually ship in. You browse, pick one, and run a single CLI command that downloads the problem, starter files, and tests, inits a git repo, and starts the timer. No sandbox, no artificial constraints—just your own editor and agent.
Three-axis scoring with evidence
Your session is scored on Direction (how well you steer the agent), Outcome (what actually shipped), and Lift (how much the agent improved over your baseline). Each axis lands with per-signal evidence, so you can see exactly why you scored the way you did.
Public leaderboard and badges
Your scores place you on a public leaderboard where you can compare your vibe coding skills against other developers. Earn badges for your achievements and build a profile that shows how you engineer, not what you memorized.
Bring your own agent
Kodwai works with Claude Code, Cursor, and Codex. You choose the agent you're most comfortable with, and the platform scores your collaboration regardless of which tool you use.
What stands out
"Passing tests is not skill. Kodwai reads the whole session, so the score rewards how you drive."
A careless one-shot prompt can make the test suite go green, but Kodwai catches the difference. It reads the entire session—your prompts, the agent's transcript, your verification steps, and how you recovered when the agent went confidently wrong. This means a developer who carefully decomposes a problem and verifies each step scores higher than someone who just ships a lucky prompt.
Worth checking out if…
You're tired of LeetCode and whiteboard puzzles that an AI agent can clear in seconds. If you want a metric that actually reflects how you direct agents, catch hallucinations, and verify what ships, Kodwai gives you a score that proves your judgment. It's free, runs on your machine, and works with the agents you already use.
