# Beefy Databarn MCP server

Schema documentation for the Beefy Databarn ClickHouse warehouse (dbt project): models,
column types, descriptions, tests, model graph and column-level lineage.
Public, read-only, metadata only. It has no access to table data.

## Connect

Endpoint (MCP streamable HTTP): https://mcp-databarn.beefy.com/

Claude Code:
  claude mcp add --transport http databarn https://mcp-databarn.beefy.com/

Other MCP clients (Cursor, VS Code, ...): add a remote/HTTP server with the URL above, or
  { "mcpServers": { "databarn": { "type": "http", "url": "https://mcp-databarn.beefy.com/" } } }

## Tools

- search: Search models, sources and columns by name or description (case-insensitive, all words must match). Returns matching models with the columns that matched.
- list_models: List models, optionally filtered by layer (staging, intermediate, dimensions, marts, api, data_quality, ...) or dbt tag.
- get_model: Full documentation for a model or source: description, ClickHouse columns with types/descriptions/tests, primary key, model-level tests, parents and children.
- column_lineage: Column-level lineage. direction='upstream' shows which source columns feed `model.column` (with transformation: passthrough/aggregated/...); 'downstream' shows which columns in other models are derived from it (impact analysis). Coverage is best-effort: some columns are not traceable (e.g. macros, arrays).
- model_graph: Model-level dependency graph around `name`. direction='upstream' (parents) or 'downstream' (children).
- freshness: When the documentation artifacts were generated and how many models are indexed.

## Layers

source -> staging (stg_*) -> intermediate (int_*) -> dimensions / marts -> api (api_*)

## Try asking

- Which models hold daily revenue, and what are their columns?
- Where does api_revenue_summary.revenue_usd_30d come from?
- What breaks if I change stg_beefy_db__feebatch_harvests.treasury_amt?

## Status

92 models, 50 sources, 0 with column lineage; docs generated 2026-10-28T01:30:02.083519Z.