Your pass to private datasets.

Traditional Sharing vs DataPass

๐Ÿ“ฅ

Traditional

Share dataset โ†’ User downloads everything

โŒ You lose control of your data

๐Ÿ”

DataPass

User asks question โ†’ Gets only the answer

โœ… Data stays private on HF

MCP Does the Heavy Lifting

When a subscriber asks a question, here's what happens behind the scenes:

1 User asks: "What are the top 10 categories?"
2 DataPass validates the access token
3 LLM converts question to SQL
4 DuckDB queries parquet files on HF
5 Only the result is returned to user

The user never sees the SQL, never touches DuckDB, never accesses HF directly.

Why DataPass?

For Dataset Owners

  • ๐Ÿ”’ Keep your data private
  • โฑ๏ธ Grant time-limited access
  • ๐Ÿšซ Revoke anytime
  • ๐Ÿ’ฐ Monetize with Stripe

For Subscribers

  • ๐Ÿ’ฌ Query with SQL or plain English
  • โšก No setup - just use MCP tools
  • ๐Ÿ“Š Get answers, not giant files
  • ๐Ÿ”Œ Works with Claude, Cursor, etc.

Security Model

๐Ÿ—„๏ธ Dataset files never leave Hugging Face
๐ŸŽซ DataPass validated on every request
๐Ÿ“‰ Results capped - no SELECT * dumps

DataPass โ€” Your pass to private datasets.

Powered by Hugging Face & MCP