# kodwai Coding challenges that score how developers guide Claude Code, Cursor, or Codex. - Findings24: - Website: - Pricing: Freemium - Categories: Developer Tools, Education Tools ## What is kodwai? kodwai is a coding practice platform for developers working with AI agents. You solve timed projects on your own machine, in your usual editor, and receive a score for both the implementation and how you guided the agent. The challenges cover real engineering tasks, including APIs, frontend interfaces, data pipelines, and distributed systems. Each includes requirements that are easy to overlook. The core purpose is to practice planning, verification, and correction when collaborating with an AI coding agent. ## What can kodwai do? ### CLI challenge setup The CLI downloads the problem statement, starter files, and tests, initializes a git repository, and starts the challenge timer. ### Three-axis scoring Scores total 100 points across Direction, Outcome, and Lift. Default weights are 50/35/15; challenges with a dedicated rubric use 45/45/10. Assessment covers agent guidance, delivered code, and improvement over a solo AI. ### Evidence-linked feedback Scoring signals reference transcripts, commits, and test runs, with a confidence interval. Each scored run highlights up to six consequential moments; when mistakes occur, at least two remain in the feedback. ### Hidden requirement traps Challenges include easily missed requirements that test edge-case coverage and whether developers question the agent's implementation. ### Developer profiles and competition Public profiles, a leaderboard, weekly leagues, and tiers provide ways to share results and participate in ongoing competition. ## Use Cases ### Practice backend interview tasks Use the 60-minute Bookshelf REST API challenge to build CRUD endpoints, combined filters, validation, persistence, and regression tests with an agent. ### Exercise frontend engineering judgment Build a searchable product listing with multi-select filters and sorting while practicing component composition, derived state, and accessible interactions. ### Test applied AI engineering skills Attempt the 120-minute Production RAG Service challenge to work through tenant isolation, cost budgets, semantic caching, idempotent ingestion, and an evaluation harness. ## Frequently Asked Questions ### What do I need to start? Create a free account and install Node.js 20+ and git. Use your own Claude Code, Cursor, or Codex agent. Run npx @kodwai/cli@latest challenge \; the CLI signs you in through the browser if needed. ### What is submitted for scoring? Running npx @kodwai/cli@latest submit packages your code, git history, test runs, agent transcript, and elapsed time, then sends them for scoring. ### Is kodwai free? Challenges, scores, profiles, and the leaderboard are free for developers. You use the AI coding agent you already pay for. ### When can I read the full challenge specification? The full problem statement becomes available when you start the challenge. Public challenge pages show the topic, difficulty, time limit, and a summary beforehand. ## Quick Facts - **Category**: AI coding practice and assessment - **Environment**: Local machine, terminal, and editor - **Challenge catalog**: 15 challenges: 3 easy, 3 medium, 9 hard - **Supported agents**: Claude Code, Cursor, Codex - **Website**: https://www.kodwai.com