概要
What is dbt-docs-mcp?
dbt-docs-mcp is an MCP server that exposes dbt project metadata from manifest.json and catalog.json artifacts. It lets you query graph information, inspect node details, and explore model and column lineage within a dbt project.
How to use dbt-docs-mcp?
Requires Python and uv. Clone the repo, optionally run the provided script to generate column-level lineage data, then start the server with uv run … mcp run … and set the environment variables MANIFEST_PATH, SCHEMA_MAPPING_PATH, and MANIFEST_CL_PATH. Configuration is typically placed in an MCP client’s mcp.json.
Key features of dbt-docs-mcp
- Search nodes by name, column name, or compiled SQL.
- Retrieve detailed attributes of any node by unique ID.
- Find direct upstream predecessors of a node.
- Find direct downstream successors of a node.
- Trace column-level ancestors and descendants.
- Supports extensions for SQL execution and database metadata.
Use cases of dbt-docs-mcp
- Discover which models depend on a given source table.
- Understand the full upstream lineage of a specific column.
- Find models whose compiled SQL contains a known pattern.
- Validate the impact of schema changes on downstream nodes.
FAQ from dbt-docs-mcp
—
