Data Platform

data-platform

Data engineering toolkit for Claude Code: pandas, PostgreSQL/PostGIS, and dbt — wired up via an MCP server that gives Claude real tools, not just instructions.

What it does

• Read and profile CSV / Parquet / pandas DataFrames
• Query PostgreSQL (and PostGIS for geospatial) with connection + column helpers
• Compile, parse, and lineage-analyze dbt projects
• Run data-quality gates

Commands

Command	Purpose
/data setup	Initialize project config
/data ingest	Ingest CSV/Parquet
/data profile	Profile a dataset (nulls, uniqueness, types)
/data schema	Extract dataset schema
/data explain	Explain SQL / pandas code
/data quality	Quality gates
/data gate	Hard-block on quality issues
/data review	Review a data pipeline
/data run	Run a dbt / pandas / SQL job
/data lineage, /data lineage-viz	Dataset lineage and visualization
/data dbt-test	dbt test with grouped output

Agents (3)

Agent	Model	Role
data-advisor	sonnet	End-to-end data pipeline advice
data-analysis	sonnet (plan mode)	Read-only dataset analysis
data-ingestion	haiku	Fast ingestion scaffolding

MCP server

mcp-servers/data-platform/ — real tools (read_csv, read_parquet, pg_connect, pg_columns, st_tables, dbt_parse, dbt_compile, dbt_lineage, etc.). The MCP is what makes this plugin worth having — without it, the commands would be prescriptive text.

Configuration

POSTGRES_URL in ~/.config/claude/postgres.env if you want PG tools; pandas/dbt tools work without it.