DBT Manifest MCP Server

A FastMCP server for analyzing DBT manifests with automatic schema version detection and lineage tracking.

Author: David B Company: DABBLEFISH LLC License: MIT

Features

• Automatic Schema Version Detection: Supports DBT manifest schema versions v0-v12

• Version-Adaptive Parsing: Backward compatibility with legacy manifest formats

• SQLite Database Storage: Efficient querying and data persistence

• Lineage Analysis: Upstream and downstream dependency tracking

• Model Information: Detailed model metadata and compiled code access

• PEP-8 Compliant: Professional Python package structure

Related MCP server: dbt-assay

Installation

From PyPI (when published)

pip install dbt-manifest-mcp

From Source

git clone https://github.com/dabblefish/dbt-manifest-mcp.git
cd dbt-manifest-mcp
pip install -e .

Development Installation

git clone https://github.com/dabblefish/dbt-manifest-mcp.git
cd dbt-manifest-mcp
pip install -e ".[dev]"

Usage

Running the Server

# Using the installed command
dbt-manifest-mcp

# Or using Python module
python -m dbt_manifest_mcp.server

Environment Variables

• DBT_MANIFEST_PATH: Path to the DBT manifest.json file (required)

• DBT_DB_PATH: Path to SQLite database file (optional, defaults to ./dbt_manifest.db)

Example

export DBT_MANIFEST_PATH="/path/to/your/manifest.json"
export DBT_DB_PATH="./dbt_manifest.db"
dbt-manifest-mcp

Available Tools

1. refresh_manifest

Refresh DBT manifest data by parsing and storing in SQLite database.

Parameters:

• manifest_path (optional): Path to the DBT manifest.json file

Returns: Success message with statistics

2. get_upstream_lineage

Get upstream lineage for a DBT model.

Parameters:

• model_id: Unique ID of the DBT model (e.g., 'model.my_project.my_model')

Returns: Dictionary with model_id, upstream_models list, and count

3. get_downstream_lineage

Get downstream lineage for a DBT model.

Parameters:

• model_id: Unique ID of the DBT model (e.g., 'model.my_project.my_model')

Returns: Dictionary with model_id, downstream_models list, and count

4. get_model_info

Get detailed information about a DBT model including parent/child counts and compiled code.

Parameters:

• model_id: Unique ID of the DBT model (e.g., 'model.my_project.my_model')

Returns: Dictionary with detailed model information

5. get_schema_info

Get information about the loaded DBT manifest schema version, supported features, and database statistics.

Returns: Dictionary with version info, features, and statistics

Schema Version Support

The server automatically detects and adapts to different DBT manifest schema versions:

• v0-v3: Legacy format with basic node structure

• v4+: Modern format with parent_map and child_map

• v12: Latest format with enhanced metadata

Version-Specific Features

Database Schema

The server creates the following SQLite tables:

• metadata: Schema version and manifest metadata

• nodes: DBT models, tests, and other nodes

• sources: DBT source definitions

• macros: DBT macro definitions

• parent_map: Parent-child relationships

• child_map: Child-parent relationships

Example Usage

# After starting the server, you can use the tools via MCP client

# Refresh manifest data
refresh_manifest("/path/to/manifest.json")

# Get upstream dependencies
upstream = get_upstream_lineage("model.my_project.customer_orders")

# Get downstream dependencies
downstream = get_downstream_lineage("model.my_project.raw_customers")

# Get detailed model information
model_info = get_model_info("model.my_project.customer_summary")

# Get schema version information
schema_info = get_schema_info()

Error Handling

The server includes comprehensive error handling for:

• Missing or invalid manifest files

• Unsupported schema versions

• Database connection issues

• Invalid model IDs

License

MIT License - see LICENSE file for details.