Claude Code Session Analyzer

Upload your Claude Code session files and receive a behavioral assessment across 6 dimensions — benchmarked against Anthropic's internal engineering cohort. Free, no sign-in required.

Drop .jsonl files or click to select

~/.claude/projects/[your-project]/*.jsonl

Privacy: Files are processed in-browser. Only anonymized session summaries (first 2 + last user message, capped at 1200 chars) are sent for AI classification. No full conversation content, code, or file paths are transmitted. Your scored report is saved and accessible via a private link.

Use Cmd+Shift+G in the file picker to navigate to ~/.claude/projects/

For hiring teams

A stronger signal than any take-home assignment

Traditional technical screens test whether a candidate can write code under artificial time pressure. They don't show how someone builds with AI across real projects: what they delegate, how much they trust the model, or whether they catch bad output before it ships.

The maturity score reads six behavioral dimensions from actual Claude Code session data. It shows the real working pattern, not the rehearsed description.

Which task types they actually delegate to AI
How much autonomy they grant across sessions
Their correction and oversight rate
Whether task complexity has grown over time

For builders

Know where you stand before the interview

Anthropic analyzed 132 of its own engineers across 200,000 Claude Code sessions and published the behavioral patterns that distinguish high performers. This tool runs the same six-dimension analysis on your sessions.

You'll see exactly where you benchmark against that cohort and where your gaps are. The report is yours to keep or share with a link.