DataFrey
Vibecode SQL. Ask your database from Claude Code.
2 followers
Vibecode SQL. Ask your database from Claude Code.
2 followers
DataFrey connects your database to AI. Ask your database in natural language with /db skill for Claude Code: "/db top 10 customers by revenue". Install the CLI, connect Snowflake, and get an MCP server for your database. A planning agent picks the right tables and joins for ambiguous questions.
Hi everyone!
I'm a data scientist. Claude Code kept writing confident SQL against schemas it didn't understand — wrong tables, invented columns, no clue what the values in a column actually looked like. Existing Snowflake MCP servers only gave it the schema. I wanted one that actually understands the data.
DataFrey is an MCP server for your database, plus a Claude Code plugin (MCP + /db skill). Snowflake only for now.
Setup is a CLI wizard — pip install datafrey && datafrey — that takes your credentials, and installs the plugin. Then in Claude you ask things like:
/db top 10 customers by revenue /db write dbt staging model for tickets
/db which customer segments are driving the revenue this quarter
For blurry questions there's a plan tool — a separate text-to-SQL agent that reads from a rich, prebuilt map of your database — schema plus the context around it — spawns subagents to poke at your data, and returns a plan before running anything.
Would love to hear what you think!