Claude MCP Data Engineer Server
A Model Context Protocol (MCP) server built for data engineers — gives Claude powerful tools to help with SQL, DBT, Snowflake, CSV analysis, pipeline validation, and ETL lineage.
Tools Available
| Tool | Description |
|---|---|
format_sql | Format & prettify SQL queries with proper indentation |
json_to_ddl | Generate CREATE TABLE DDL from a JSON sample record |
analyze_csv | Analyze CSV data — row count, column types, null counts |
generate_dbt_model | Generate a dbt model SQL file with source + deduplication CTE |
validate_pipeline_config | Validate a data pipeline JSON config for required fields |
generate_snowflake_table | Generate a Snowflake CREATE TABLE with optional CLUSTER BY |
summarize_etl_lineage | Summarize ETL lineage JSON into a human-readable pipeline flow |
Setup
1. Install dependencies
pip install -r requirements.txt
2. Run the MCP server
python -m src.server
3. Connect to Claude Desktop
Add this to your claude_desktop_config.json (~/AppData/Roaming/Claude/ on Windows):
{
"mcpServers": {
"data-engineer": {
"command": "python",
"args": ["-m", "src.server"],
"cwd": "C:/Users/Nikhil/claude-mcp-data-engineer"
}
}
}
Example Usage
Format SQL
Ask Claude:
"Format this SQL: select id,name from users where status='active' group by id"
Generate DDL from JSON
Ask Claude:
"Generate a CREATE TABLE DDL for this JSON:
{"id": 1, "name": "Nikhil", "salary": 95000.0, "joined": "2024-01-15"}"
Analyze CSV
Ask Claude:
"Analyze this CSV data and tell me column types and null counts"
Generate DBT Model
Ask Claude:
"Generate a dbt model for source 'raw', table 'orders', columns: id, customer_id, amount, created_at"
Validate Pipeline Config
Ask Claude:
"Validate this pipeline config:
{"source": {"type": "snowflake"}, "destination": {"type": "s3"}, "schedule": "0 2 * * *"}"
Tech Stack
- Python 3.10+
- MCP SDK (
mcp[cli]) - FastMCP for server definition
Author
Nikhil E — Sr. Data Engineer | BI Architect GitHub: itsnikhile