Kodwai is a coding challenge platform where you solve real-world problems on your machine using AI agents like Claude Code, Cursor, and Codex, with leaderboards and a professional portfolio.

What AI agents does kodwai support?

Kodwai supports AI agents such as Claude Code, Cursor, and Codex to help you solve coding challenges.

How does kodwai work?

You solve real-world coding problems on your own machine with the help of AI agents, then submit your solutions to compete on leaderboards and build your portfolio.

Is kodwai free to use?

Kodwai offers free access to basic features, but premium plans may be available for advanced challenges and leaderboards.

Can I use kodwai to improve my coding skills?

Yes, kodwai helps you improve by solving practical problems with AI assistance, tracking progress on leaderboards, and building a professional portfolio.

How do I get started with kodwai?

Sign up at kodwai.com, choose a challenge, and start solving it on your machine using supported AI agents.

kodwai - AI Code generation tools - Free trial, pricing intro, performance review, official site access and online experience

What is kodwai?

Kodwai is a coding challenge platform where developers solve real-world problems on their own machine using AI agents like Claude Code, Cursor, or Codex. Instead of testing memorized algorithms, it scores how well you direct an agent—catching hallucinations, verifying outputs, and shipping real code. The platform ranks you on a public leaderboard based on your ability to manage AI-driven development. It’s designed to measure the skill that matters in modern engineering: directing an agent effectively, not just passing tests.

Application scenarios

AI agent skill assessment
Developers test their ability to prompt, verify, and recover from agent errors in real coding tasks.
Team hiring or internal evaluation
Engineering teams can use Kodwai to evaluate candidates or current staff on agent-directed problem-solving.
Competitive coding with AI
Compete on leaderboards by solving challenges with your own agent, comparing scores on direction, outcome, and lift.
Skill benchmarking
See how your agent-handling skills rank against other developers in a public, transparent way.
Learning and improvement
Review per-signal evidence (direction, outcome, lift) to understand why you scored the way you did and improve.
Real-world workflow simulation
Work on ticket-sized problems in your own editor, with no sandbox constraints, mimicking how you actually build software.

Core Features

Bring your own agent
Use Claude Code, Cursor, or Codex—no proprietary AI required; you work with the agent you already use.
Five-step challenge flow
Pick a challenge, run a CLI command, solve on your machine, submit with one command, and get scored.
CLI-based submission
Run `$npx @kodwai/cli submit` to package code, git history, test runs, agent transcript, and time for scoring.
Three-score evaluation
You are scored on Direction (how well you direct the agent), Outcome (what shipped), and Lift (improvement over baseline).
Per-signal evidence
Each score axis lands with specific evidence so you can see why you scored that way.
Public leaderboard
Rankings are visible to all, showing how your agent-handling skill compares.
Real-world problems
Challenges are ticket-sized, covering categories you actually ship in, with difficulty filters.
No sandbox constraints
Work in your own editor and terminal with your own agent—no artificial limitations.
Git history tracking
The CLI inits a git repo and tracks your session, including agent transcript and test runs.
Free to start
Challenges are fully free; you bring your own agent (Claude Code, Cursor, or Codex).

Target users

Kodwai is built for developers who work with AI agents daily—engineers who want to prove their skill at directing agents, not just memorizing LeetCode patterns. It’s also useful for engineering managers hiring for modern roles, and for teams evaluating how well members handle agent-driven development. The platform targets anyone who believes the real skill is in catching agent errors, writing specs, and shipping verified code.

How to use kodwai?

Pick a challenge: Browse real, ticket-sized problems on the Kodwai website, filter by difficulty, and choose one.
Run the CLI: In your terminal, run $npx @kodwai/cli challenge to download PROBLEM.md, starter files, and tests. Choose your agent (Claude Code, Cursor, or Codex).
Solve on your machine: Work the problem in your own editor with your own agent. No sandbox—just how you really build.
Submit: Run $npx @kodwai/cli submit to package your code, git history, test runs, agent transcript, and time for scoring.
Get scored: View your Direction, Outcome, and Lift scores with per-signal evidence, then see your rank on the leaderboard.

Pricing and free trial

Kodwai is fully free to start. The website states: "Start a challenge→fully free/bring your own agent/claude code, cursor or codex." There is no mention of paid tiers or trial limitations in the provided text.

Effect review

Kodwai addresses a genuine gap in developer assessment—LeetCode-style tests don’t measure how well you handle AI agents in real workflows. The three-score system (Direction, Outcome, Lift) with per-signal evidence provides transparent, actionable feedback, which is rare in coding challenge platforms. The CLI-based flow is straightforward and respects how developers actually work, avoiding sandbox friction. However, the platform’s value depends entirely on the quality and relevance of its challenge library, which isn’t detailed in the text. For developers already using AI agents daily, Kodwai offers a practical, competitive way to benchmark a skill that’s increasingly central to modern engineering.

kodwai