Six months of Claude Code sessions. One honest summary. Stays on your laptop.
Your ~/.claude/projects/ folder is keeping a log on you.
Every conversation with Claude Code lands as a JSONL file you've never opened. After a few months, that's gigabytes of receipts on how you actually code. Crixin reads it locally, dumps it into a DuckDB, and gives you a dashboard you can read in 30 seconds.
Concrete numbers, not vibes. Each one comes from your real Claude Code history — not telemetry, not someone else's dataset.
You'll see the patterns you didn't realize you had. The "fix this regex" loop, the "explain this stack trace" loop, the "rename this variable across the codebase" loop.
Which project is eating your token budget? Which models do you actually reach for at 2am vs 2pm? The dashboard answers in one screen.
Every search result includes a ready-to-paste claude --resume <uuid> command. The auth refactor from three weeks ago is one copy-paste away.
Crixin reads ~/.claude/projects/<project>/<session-uuid>.jsonl directly. No instrumentation. No middleware.
# Default: ingest everything in ~/.claude/projects, open the dashboard. $ npx crixin # Find sessions by content (sub-second). $ npx crixin search "auth refactor" # Resume one of them. $ cd ~/Documents/myapp && claude --resume 7897c935-2069-4b75-bbad-a3fac62ea59c # Drop a session as Markdown into your notes. $ npx crixin export 7897c935-... > notes/auth-refactor.md
No account. No upload. No telemetry. The dashboard runs on 127.0.0.1 and binds to a random port.
— nothing —
— nothing —
Free for life on Claude Code. $5/mo for the pro packs (cost analysis, eval scores, prompt diffs).