Data Product MCP Server

An MCP server designed to discover and query Data Products on any data platform in a governed way - enabling AI-assistants to answer any business question.

Overview

The Data Product MCP Server provides an interface for AI clients to discover and query data products via MCP. It enables AI agents to locate, analyze, and execute queries data to answer business questions in natural language on all available business data products.

Example

What was the revenue in our webshop yesterday? What about cancellations? What were the reasons for them?

Key Features

• Discovery: Find relevant data products. Data contracts provide the necessary schema and semantic context information.
• Governance Enforcement: Automate access management and checks for terms of use.
• Data Querying: Natural language questions are translated into SQL, executed, and augumented.
• Federated Queries (Alpha): Join data across multiple data products from different sources and technologies.
• Local & Remote Support: Work with local data product repositories or remote data product marketplaces, such as Data Mesh Manager
• Plugin Architecture: Extensible plugin system for adding new asset sources and data sources

Use Cases

• AI assistants exploring available data products
• Analyzing data contract schemas and requirements
• Executing SQL queries against data products
• Joining data across multiple products for complex analyses
• Validating data quality against contract specifications
• Building data-aware AI applications

Installation

Prerequisites

• Python 3.11 or higher
• uv (recommended) or pip

Quick Install

Using uv (recommended):

uv pip install -e .

Using pip:

pip install -e .

Running the Server

Basic Usage

# Set location of data files
export DATAASSET_SOURCE=/path/to/assets/directory

# Start the server
python -m src.dataproduct_mcp.server

Development Mode

# For local development with MCP CLI
mcp dev -m src.dataproduct_mcp.server

# For debugging (runs test query)
python -m src.dataproduct_mcp.server

# For server mode
python -m src.dataproduct_mcp.server --server

Integration with Claude Code or Claude Desktop

Add this configuration to your Claude installation:

{
  "mcpServers": {
    "dataproduct": {
      "command": "uv",
      "args": ["run", "--directory", "<path_to_folder>/dataproduct-mcp", "python", "-m", "dataproduct_mcp.server"],
      "env": {
        "DATAASSET_SOURCE": "<path_to_folder>/dataproduct-mcp/examples"
      }
    }
  }
}

Tools and Capabilities

Data Product Operations

Tool	Description
dataproducts_list	List all available data products
dataproducts_get	Retrieve a specific data product by identifier
dataproducts_get_output_schema	Get schema for a data contract linked to a product output port
dataproducts_query	Execute SQL queries against one or multiple data product output ports

Examples

List Available Data Products

dataproducts_list

Get a Specific Data Product

dataproducts_get(identifier="local:product/orders.dataproduct.yaml")

dataproducts_get(identifier="datameshmanager:product/customers")

Get Schema for a Data Contract

dataproducts_get_output_schema(identifier="local:contract/orders.datacontract.yaml")

dataproducts_get_output_schema(identifier="datameshmanager:contract/snowflake_customers_latest_npii_v1")

Query a Single Data Product

dataproducts_query(
    sources=[
        {
            "product_id": "local:product/orders.dataproduct.yaml"
        }
    ],
    query="SELECT * FROM orders LIMIT 10;",
    include_metadata=True
)

Execute a Federated Query (Alpha)

dataproducts_query(
    sources=[
        {
            "product_id": "local:product/orders.dataproduct.yaml",
            "alias": "orders"
        },
        {
            "product_id": "local:product/video_history.dataproduct.yaml", 
            "alias": "videos"
        }
    ],
    query="SELECT o.customer_id, v.video_id, o.order_date, v.view_date FROM orders o JOIN videos v ON o.customer_id = v.user_id WHERE o.order_date > '2023-01-01'",
    include_metadata=True
)

Note: Querying multiple data products at once is still in alpha status and may have limitations.

Included Examples

The examples directory contains sample files for both data products and their supporting data contracts:

Sample Data Products

• orders.dataproduct.yaml - Customer order data
• shelf_warmers.dataproduct.yaml - Products with no sales in 3+ months
• video_history.dataproduct.yaml - Video viewing history data

Sample Data Contracts

• orders.datacontract.yaml - Defines order data structure and rules
• shelf_warmers.datacontract.yaml - Contract for product inventory analysis
• video_history.datacontract.yaml - Specifications for video consumption data

Architecture

The Data Contract MCP Server uses a plugin-based architecture that cleanly separates:

Asset Sources (Metadata)

These plugins handle loading and listing data assets (products and contracts):

• Local files: Loads YAML files from a local directory
• Data Mesh Manager: Fetches assets from a Data Mesh Manager API

Data Sources (Query Execution)

These plugins handle the actual data querying:

• Local files: Queries CSV, JSON, or Parquet files using DuckDB
• S3: Queries data in S3 buckets using DuckDB's S3 integration
• Databricks: Queries data from Databricks SQL warehouses

The plugin architecture makes it easy to add support for additional asset sources (e.g., Git repositories, other APIs) and data sources (e.g., databases, data warehouses).

Configuration Options

Environment Variables

Variable	Description	Default
DATAASSET_SOURCE	Directory containing data assets	Current directory
DATAMESH_MANAGER_API_KEY	API key for Data Mesh Manager	None
DATAMESH_MANAGER_HOST	Host URL for Data Mesh Manager	https://api.datamesh-manager.com

AWS S3 Configuration (for S3 data sources)

• AWS_REGION / AWS_DEFAULT_REGION - AWS region (default: us-east-1)
• S3_BUCKETS - Allowed S3 buckets (comma-separated)
• Authentication via profile (AWS_PROFILE) or credentials (AWS_ACCESS_KEY_ID/AWS_SECRET_ACCESS_KEY)

Databricks Configuration (for Databricks data sources)

• DATABRICKS_WORKSPACE_URL - Databricks workspace URL (required)
• DATABRICKS_TOKEN - Personal access token for Databricks
• DATABRICKS_CLIENT_ID / DATABRICKS_CLIENT_SECRET - OAuth client credentials (alternative to token)
• DATABRICKS_CATALOG - Default catalog to use (optional)
• DATABRICKS_SCHEMA - Default schema to use (optional)
• DATABRICKS_TIMEOUT - Query execution timeout in seconds (default: 120)

Development Setup

Python base interpreter should be 3.11.x.

# create venv
python3.11 -m venv venv
source venv/bin/activate

# Install Requirements
pip install --upgrade pip setuptools wheel
pip install -e '.[dev]'
pre-commit install
pre-commit run --all-files
pytest

Use uv (recommended)

# make sure uv is installed
uv python pin 3.11
uv pip install -e '.[dev]'
uv run ruff check
uv run pytest

Contribution

We are happy to receive your contributions. Propose your change in an issue or directly create a pull request with your improvements.

Related Tools

• Data Contract CLI is an open-source command-line tool for working with data contracts.
• Data Contract Manager is a commercial tool to manage data contracts. It contains a web UI, access management, and data governance for a full enterprise data marketplace.
• Data Contract GPT is a custom GPT that can help you write data contracts.
• Data Contract Editor is an editor for Data Contracts, including a live html preview.
• Data Contract Playground allows you to validate and export your data contract to different formats within your browser.

License

MIT License