Reveal your Gradient Score

Drop in your local usage files and we'll measure how well you work with AI — right here in your browser.

This measures how you use Claude Code (CLI) — read locally from your /insights usage data, never uploaded.

Easiest: create the file with one command

Run this where you actually use Claude Code — in your computer's terminal. It won't work in the claude.ai web chat (that's a sandbox without your files). Your data stays on your machine; you only upload the file it creates.

Run in your terminal

python3 -c "import json,glob,os; d=lambda p:[json.load(open(f)) for f in sorted(glob.glob(os.path.expanduser(p)))]; json.dump({'facets':d('~/.claude/usage-data/facets/*.json'),'sessionMeta':d('~/.claude/usage-data/session-meta/*.json')}, open('gradient-export.json','w')); print('wrote gradient-export.json')"

Prefer to ask Claude Code? Paste this prompt instead:

In Claude Code on my own computer: read every JSON file in ~/.claude/usage-data/facets/ and ~/.claude/usage-data/session-meta/. Create a single file "gradient-export.json" in the current directory with exactly this shape: {"facets":[ ...each facets object... ],"sessionMeta":[ ...each session-meta object... ]}. Do not read or include anything else. Then print the path to gradient-export.json.

Select your usage files

The JSON files from ~/.claude/usage-data/facets and ~/.claude/usage-data/session-meta.

or drag & drop them here

Your files are read in your browser. Nothing is uploaded.