The interview platform
that sees how engineers
actually work.
The first platform where candidates use Claude Code — a real AI coding agent, not a chatbot. Full session capture. AI-powered scoring.
Free early access · No card required
Built for teams at
The Problem
You're testing skills
that don't matter anymore.
Whiteboard algorithms. Toy problems. No AI allowed. These interviews measure memorization, not engineering. The best engineers ship with agents — your interview should test that.
Real agents, not chatbots
Candidates use Claude Code — a full autonomous agent with file system, terminal, and multi-file project access. Not a prompt box.
Signal, not noise
Every keystroke, architectural decision, and agent interaction is captured and scored. You see exactly how engineers think.
Built for the AI era
The old whiteboard is dead. kodwai tests the skill that actually matters: wielding AI agents to ship production code.
How It Works
Three steps. Total clarity.
Deploy a challenge
Pick from our library or create custom system-design and coding challenges calibrated to your stack.
Candidate solves with AI
They get a full Claude Code environment. Real agent, real tools, real constraints. 60 minutes.
Review the signal
AI-generated scorecard with granular breakdowns: decomposition, agent mastery, code quality, verification.
Comparison
The data sheet.
“We stopped asking candidates to reverse linked lists. Now we see how they actually build with AI.”
— VP Engineering, Series B startup
Hiring signal accuracy
Faster than traditional loops
Engineers on the waitlist
Stop guessing.
Join the waitlist. Be first to interview engineers the way they actually work.