getsit

Gets It

A Claude Code plugin that measures how well your AI actually gets you.

Gets It silently tracks decision points in your conversations — moments where you're choosing between options — and measures how accurately the LLM predicts what you'll pick. Over time, it builds a personalization score across four domains: work, personal, technical, and creative.

Install

1. Add the marketplace

/plugin marketplace add andrewsiah/getsit

2. Install the plugin

/plugin install getsit@andrewsiah-getsit

When prompted, select User scope to activate it across all projects.

That's it. The plugin runs silently in the background from now on — no flags, no configuration.

Alternative: Install from local path (for development)

/plugin install getsit --path /path/to/getsit

How It Works

1. You have a normal conversation with Claude
2. When a decision point comes up (e.g., "Should I use Postgres or SQLite?"), the plugin silently predicts your choice
3. After you decide, it resolves the prediction
4. Your accuracy score builds over time

You won't notice it running. That's by design — unbiased predictions require that you don't know when one is being made.

View Your Score

Ask Claude:

"Show me my Gets It dashboard"

Or run the dashboard directly:

python3 /path/to/getsit/scripts/dashboard.py

====================================================
  Gets It Dashboard
====================================================

  Overall:   72.0%  (18/25)
  Last 7d:   80.0%  (8/10)

  By Domain:
  --------------------------------------------
  personal     [################....] 80.0%  (4/5)
  technical    [##############......] 70.0%  (7/10)
  work         [###############.....] 75.0%  (6/8)
  creative     [##########..........] 50.0%  (1/2)

  Recent:
  --------------------------------------------
  [+] technical    Chose Postgres over SQLite
  [-] work         Picked async over sync standup
  [+] personal     Went with morning routine
  [+] technical    REST over GraphQL
  [+] work         Monorepo over multi-repo

====================================================

Add --json for machine-readable output.

What Gets Tracked

Domain	Examples
work	Project decisions, prioritization, process
personal	Lifestyle, preferences, daily choices
technical	Architecture, tooling, languages, frameworks
creative	Design, naming, branding, aesthetics

Data Storage

All data is local. Predictions are stored in ~/.getsit/predictions.jsonl — a plain-text file you can inspect, edit, or delete at any time.

No data leaves your machine. No accounts. No telemetry.

Requirements

• Claude Code CLI
• Python 3.10+ (for the dashboard script — uses stdlib only, no pip install needed)

Uninstall

/plugin disable getsit@andrewsiah-getsit

Or remove it entirely:

/plugin uninstall getsit@andrewsiah-getsit

Your prediction data stays in ~/.getsit/ until you delete it yourself.

License

MIT