kodwai
Scores how well you direct an AI coding agent through real challenges, not just tests.
kodwai gives you real coding challenges to solve on your own machine with your own AI agent — Claude Code, Cursor or Codex — then scores the session on three axes: Direction (how well you steered the agent), Rubric (correctness against hidden tests and traps), and Lift (how much you added beyond the agent's first pass). Its own example shows why that matters: the same challenge, same agent, scored 44 of 100 when a developer pasted the spec and walked away, versus 90 of 100 when they wrote a spec, added tests and pushed back twice. It's free for developers and runs via a CLI (npx @kodwai/cli).
What holds up
- +Scores how you directed the agent (Direction, Rubric, Lift), not just whether tests passed.
- +Runs on your own machine with the AI agent and editor you already use and pay for.
- +Free for developers, with no browser sandbox to fight.
Mind the limits
- −Needs your own paid AI agent subscription (Claude Code, Cursor or Codex) — kodwai provides none.
- −Challenges are timed (e.g. 110 minutes) with hidden traps — real pressure, not open-ended practice.
Picked here? Add the badge
Put it on your site. It links back to this page. Free for every listed product.
<a href="https://stillworks.dev/products/p/kodwai/"><img src="https://stillworks.dev/badge/kodwai.svg" alt="Picked by StillWorks" width="230" height="54"></a>