Product & Vision · 2026
kodwai
Where developers prove how they work with AI agents, and companies hire for it.
Every company now expects engineers to ship with AI agents. No platform measures that skill, let alone teaches it.
This is the wedge.
Two platforms. One skill.
Developer-first. Business-funded.
Developer Platform · Free
Solve real, ticket-sized challenges with Claude Code, Cursor, or Codex, on your own machine. Get scored on how well you direct it. Climb a public leaderboard.
Business Platform · Paid
Interview on deeper, role-specific challenges, or author your own. Watch the real process, not just the answer, then score and compare candidates.
The Developer Problem
The Gap
How skill is tested today
How engineers actually work
The Developer Platform, Free and Live
kodwai
Pick a challenge, solve it on your own machine with Claude Code, Cursor, or Codex, and submit the session. Kodwai captures every prompt, tool call, and decision, then scores how well you directed the agent. Climb a public leaderboard and build a profile that shows how you engineer, not what you memorized. Fully free.
How It Works, Live Today
Three steps. One hour. Complete signal.
Pick a challenge
Browse real, ticket-sized problems across nine categories. Filter by difficulty and pick one that looks like the work you actually do.
Run the CLI
Start it in your terminal and choose Claude Code, Cursor, or Codex. Kodwai pulls the problem, starter files, and tests, initializes git, and starts the timer. You solve it for real.
Get your score
On submit, Kodwai reads your transcript, code, tests, and git history, then returns Direction, Outcome, and Lift with per-signal evidence. Then you are on the leaderboard.
The AI Collaboration Score
Not a pass/fail. A complete picture.
The skill the free platform measures is the skill companies are desperate to hire for. So we sell it to them.
Act II: the business.
The Business Platform, Built Today
Hiring, measured.
Interview on deeper, role-specific challenges, or author your own with a custom rubric and time limit. Invite a candidate, they solve on their own machine with Claude Code or Cursor, and you watch the real process live: the prompts, the commits, the tests, the score. Not just the final answer.
Market Opportunity
A massive market with an empty quadrant.
Why Now
The signals are loud.
Business Model
Recurring revenue. Zero LLM cost.
Developer · the funnel
Business · the revenue
From Free Audience to Revenue
Free developers. Paid companies.
A trusted, free developer audience gives us reach and credibility. We sell the hiring product to the companies those developers already work at.
The leaderboard surfaces developers who provably direct agents well. That becomes a vetted talent signal companies pay to access.
Developers who practice for free get funneled into employer-run interviews on the same engine. The proven LeetCode and HackerRank playbook.
Individual engineers and teams adopt Kodwai, then expand into org-wide paid hiring. The buyer, the company, differs from the free user.
What's Built
Full platform. Shipping today.
- ▶npx @kodwai/cli challenge <slug>: one-command launch
- ▶Captures every prompt, tool call, and file edit via hooks
- ▶Real-time file and transcript watchers
- ▶HMAC-SHA256 signed event streaming
- ▶Auto-installs Claude Code, timer, auto-submit
- ▶Developer: challenges, leaderboard, profiles, badges
- ▶Hiring: projects, rubric builder, tool constraints
- ▶Live observer: transcript, diffs, tool feed, timer
- ▶Dual scoring and multi-candidate comparison
- ▶Team management with role-based access
- ▶AI Collaboration Score: 3 axes, 11 signals
- ▶Full session lifecycle with HMAC webhooks
- ▶AES-256-GCM encrypted keys, JWT auth
- ▶Per-session cost tracking via proxy
- ▶Transactional email via Resend
The Path
Roadmap
Team
Built by people who live this problem.
A two-person team that shipped the CLI, web app, API, and scoring engine before taking a dollar.
Competition
The white space is real.
| Own-terminal agent | Free dev community | Agent-skill native | |
|---|---|---|---|
| HackerRank $221M rev · 26M devs | ✗ | ✓ | ✗ |
| CodeSignal ~$90M raised | ✗ | ✗ | ✓ |
| CoderPad ~$18M rev | ✗ | ✗ | ✗ |
| Codility 12M+ assessments | ✗ | ✗ | ✗ |
| LeetCode 26M visitors | ✗ | ✓ | ✗ |
| kodwai | ✓ | ✓ | ✓ |
Incumbents capture AI inside their own browser sandbox, and the ones with a community built it for puzzle prep. None pair a real agent in the developer's own terminal with a free developer-led community and hiring monetization. To our knowledge, as of mid-2026, that intersection is empty.
Funding and revenue figures are public reporting, 2024 to 2026.
The Agent SDK is public. The moat is the community and the dataset: real, labeled agent-collaboration data that no incumbent can buy.
Category ownership, won in a 12 to 18 month window.
Where developers prove it, and companies hire it.
Developers prove how they build. Companies hire who can. One platform, one skill, the new standard for the AI-coding era.