Now in Beta

Hire engineers who
actually know how
to use AI

Vetted evaluates real Claude Code sessions to measure craft, judgment, and technical fluency — so you hire based on signal, not theater.

Free during beta · No credit card required

Resumes can't measure AI fluency

Traditional hiring signals break down when the most important skill is how someone collaborates with AI.

Resumes list tools, not skill

Saying "proficient in Claude Code" tells you nothing about how someone actually uses it — or whether they can think alongside AI.

Coding tests miss the point

LeetCode doesn't measure whether an engineer can debug AI output, leverage tooling, or make real trade-offs under ambiguity.

Interviews are expensive guesswork

A 45-minute interview can't reveal how someone works through ambiguous problems with AI assistance over hours of real work.

How it works

Three steps to better hiring signal

Replace guesswork with evidence. Get a clear picture of how candidates actually work with AI.

Step 01

Candidate completes a session

They work on a real problem using Claude Code. No contrived puzzles — just authentic engineering work that shows how they actually think and build.

Step 02

Vetted analyzes everything

Our AI evaluates the full session across 4 dimensions — craft, clarity, judgment, and ambition — using rubrics built from real engineering standards.

Step 03

You get actionable signal

A detailed report with scores, evidence, and specific observations. Enough signal to make a confident hiring decision — in minutes, not weeks.

Evaluation rubric

Four dimensions of excellence

Each session is scored across the dimensions that matter most for AI-native engineering — with evidence and specific observations.

Scored 1–5

Craft

Tool mastery & best practices

Do they use CLAUDE.md files, MCP servers, hooks, and custom slash commands? Or are they using Claude Code like a basic chatbot?

Scored 1–5

Clarity

Communication & intent

Can they articulate what they want clearly? Do they provide context, set constraints, and guide the AI with domain knowledge?

Scored 1–5

Judgment

Decision-making & critical thinking

Do they catch AI mistakes? Push back on bad suggestions? Make thoughtful trade-offs when multiple approaches exist?

Scored 1–5

Ambition

Scope & real-world impact

Are they solving real problems with meaningful complexity? Or just generating boilerplate and calling it a day?

What you get

Reports with real signal

Not a pass/fail. A nuanced breakdown of how the candidate thinks, builds, and collaborates with AI.

vetted.so/report/abc123
Session Analysis
4.2

Strong AI-native engineer

Demonstrates advanced Claude Code proficiency with strong tool usage, clear intent communication, and solid technical judgment.

Craft4.5
Clarity4
Judgment4.5
Ambition3.8
Standout moment

“When Claude suggested a recursive approach, the candidate identified the O(2^n) complexity issue, pushed back, and guided Claude toward a dynamic programming solution with memoization — demonstrating both technical depth and effective AI collaboration.”

FAQ

Common questions

We accept .txt, .jsonl, and .json files up to 10MB. You can export sessions from Claude Code using /export, or upload the raw .jsonl files from ~/.claude/projects/.

Our AI analyzes the full session transcript against a rubric covering four dimensions: Craft, Clarity, Judgment, and Ambition. Each dimension is scored 1–5 with specific evidence from the session.

Most sessions are analyzed in under 60 seconds. You'll see real-time progress while we process your file.

Session data is used only for analysis and is never shared with third parties. We hash IPs for rate limiting and don't store personal information beyond what's needed for the analysis.

Yes — each report has a unique, shareable link. You can send it to candidates for transparency, or keep it internal.

Free during beta. We'll announce pricing well before any changes, and early users will get preferential rates.

Stop guessing. Start evaluating.

Join the waitlist and be first to hire with real AI fluency signal.