The insight: these CLIs already index your schemas, migrations, and models. Instead of building another NL2SQL tool with custom embeddings, I just wrapped what already works.

How it's different: - No schema sync. New table? git pull. Done. - No embeddings to regenerate - Follow-up queries work naturally (session persists) - Uses your actual table/column names

Usage: $ cd your-project $ qry init $ qry > get users who signed up last week SELECT * FROM users WHERE created_at >= NOW() - INTERVAL '7 days';

Also exposes an API server for Slack bots, admin tools, etc.

GitHub: https://github.com/amansingh-afk/qry

Tradeoff: requires one of these CLIs installed. But if you're already using Claude Code or Cursor, it just works.

Would love feedback on the approach.