mattijsdp/dbt-docs-mcp

Model Context Protocol (MCP) server for interacting with dbt project metadata, including dbt Docs artifacts (manifest.json, catalog.json). This server exposes dbt graph information and allows querying node details, model/column lineage, and related metadata.
Key Functionality
This server provides tools to:
* **Search dbt Nodes:**
* Find nodes (models, sources, tests, etc.) by name (search_dbt_node_names).
* Locate nodes based on column names (search_dbt_column_names).
* Search within the compiled SQL code of nodes (search_dbt_sql_code).
* **Inspect Nodes:**
* Retrieve detailed attributes for any given node unique ID (get_dbt_node_attributes).
* **Explore Lineage:**
* Find direct upstream dependencies (predecessors) of a node (get_dbt_predecessors).
* Find direct downstream dependents (successors) of a node (get_dbt_successors).
* **Column-Level Lineage:**
* Trace all upstream sources for a specific column in a model (get_column_ancestors).
* Trace all downstream dependents of a specific column in a model (get_column_descendants).
* **Suggested extensions:**
* Tool that allows executing SQL queries.
* Tool that retrieves table/view/column metadata directly from the database.
* Tool to search knowledge-base.
Getting Started
1. **Prerequisites:** Ensure you have Python installed and uv
2. **Clone the repo:**
3. **Optional: parse dbt manifest for column-level lineage:**
- Setup the required Python environment, e.g.:
- Use the provided script scripts/create_manifest_cl.py and simply provide the path to your dbt manifest, dbt catalog and the desired output paths for your schema and column lineage file:
- Depending on your dbt project size, creating column-lineage can take a while (hours)
4. **Run the Server:**
- If your desired MCP client (Claude desktop, Cursor, etc.) supports mcp.json it would look as below: