MCP Neo4j Data Modeling

🔍📊 Neo4j Data Modeling MCP Server

mcp-name: io.github.neo4j-contrib/mcp-neo4j-data-modeling

🌟 Overview

A Model Context Protocol (MCP) server implementation that provides tools for creating, visualizing, and managing Neo4j graph data models. This server enables you to define nodes, relationships, and properties to design graph database schemas that can be visualized interactively.

This MCP server facilitates data modeling workflows like the one detailed below.

• Blue steps are handled by the agent
• Purple by the Data Modeling MCP server
• Green by the user

Demo

For an end to end demo using the Data Modeling and Cypher MCP servers to develop a data model, generate an ingest script, and validate use cases please check out this Github Repo.

🧩 Components

📦 Resources

The server provides these resources:

Schema

• resource://schema/node
• Get the JSON schema for a Node object
• Returns: JSON schema defining the structure of a Node

• resource://schema/relationship
• Get the JSON schema for a Relationship object
• Returns: JSON schema defining the structure of a Relationship

• resource://schema/property
• Get the JSON schema for a Property object
• Returns: JSON schema defining the structure of a Property

• resource://schema/data_model
• Get the JSON schema for a DataModel object
• Returns: JSON schema defining the structure of a DataModel

Example Data Models

• resource://examples/patient_journey_model
• Get a real-world Patient Journey healthcare data model in JSON format
• Returns: JSON DataModel for tracking patient encounters, conditions, medications, and care plans

• resource://examples/supply_chain_model
• Get a real-world Supply Chain data model in JSON format
• Returns: JSON DataModel for tracking products, orders, inventory, and locations

• resource://examples/software_dependency_model
• Get a real-world Software Dependency Graph data model in JSON format
• Returns: JSON DataModel for software dependency tracking with security vulnerabilities, commits, and contributor analysis

• resource://examples/oil_gas_monitoring_model
• Get a real-world Oil and Gas Equipment Monitoring data model in JSON format
• Returns: JSON DataModel for industrial monitoring of oil and gas equipment, sensors, alerts, and maintenance

• resource://examples/customer_360_model
• Get a real-world Customer 360 data model in JSON format
• Returns: JSON DataModel for customer relationship management with accounts, contacts, orders, tickets, and surveys

• resource://examples/fraud_aml_model
• Get a real-world Fraud & AML data model in JSON format
• Returns: JSON DataModel for financial fraud detection and anti-money laundering with customers, transactions, alerts, and compliance

• resource://examples/health_insurance_fraud_model
• Get a real-world Health Insurance Fraud Detection data model in JSON format
• Returns: JSON DataModel for healthcare fraud detection tracking investigations, prescriptions, executions, and beneficiary relationships

Ingest

• resource://neo4j_data_ingest_process
• Get a detailed explanation of the recommended process for ingesting data into Neo4j using the data model
• Returns: Markdown document explaining the ingest process

🛠️ Tools

The server offers these core tools:

✅ Validation Tools

• validate_node
• Validate a single node structure
• Input:
• node (Node): The node to validate
• return_validated (bool, optional): If True, returns the validated node object instead of True
• Returns: True if valid (or validated Node object if return_validated=True), raises ValueError if invalid

• validate_relationship
• Validate a single relationship structure
• Input:
• relationship (Relationship): The relationship to validate
• return_validated (bool, optional): If True, returns the validated relationship object instead of True
• Returns: True if valid (or validated Relationship object if return_validated=True), raises ValueError if invalid

• validate_data_model
• Validate the entire data model structure
• Input:
• data_model (DataModel): The data model to validate
• return_validated (bool, optional): If True, returns the validated data model object instead of True
• Returns: True if valid (or validated DataModel object if return_validated=True), raises ValueError if invalid

👁️ Visualization Tools

• get_mermaid_config_str
• Generate a Mermaid diagram configuration string for the data model, suitable for visualization in tools that support Mermaid
• Input:
• data_model (DataModel): The data model to visualize
• Returns: Mermaid configuration string representing the data model

🔄 Import/Export Tools

These tools provide integration with [Arrows](https://arrows.app/) - a graph drawing web application for creating detailed Neo4j data models with an intuitive visual interface.

• load_from_arrows_json
• Load a data model from Arrows app JSON format
• Input:
• arrows_data_model_dict (dict): JSON dictionary from Arrows app export
• Returns: DataModel object

• export_to_arrows_json
• Export a data model to Arrows app JSON format
• Input:
• data_model (DataModel): The data model to export
• Returns: JSON string compatible with Arrows app

• load_from_owl_turtle
• Load a data model from OWL Turtle format
• Input:
• owl_turtle_str (str): OWL Turtle string representation of an ontology
• Returns: DataModel object with nodes and relationships extracted from the ontology
• Note: This conversion is lossy - OWL Classes become Nodes, ObjectProperties become Relationships, and DatatypeProperties become Node properties.

• export_to_owl_turtle
• Export a data model to OWL Turtle format
• Input:
• data_model (DataModel): The data model to export
• Returns: String representation of the data model in OWL Turtle format
• Note: This conversion is lossy - Relationship properties are not preserved since OWL does not support properties on ObjectProperties

• export_to_pydantic_models
• Export a data model to Pydantic models
• Input:
• data_model (DataModel): The data model to export
• Returns: String representation of the Pydantic models as a Python file, including imports and model definitions for nodes, relationships, and the complete data model

• export_to_neo4j_graphrag_pkg_schema
• Export a data model to Neo4j GraphRAG Python Package schema format
• Input:
• data_model (DataModel): The data model to export
• Returns: Dictionary containing the Neo4j GraphRAG Python Package schema

• load_from_neo4j_graphrag_pkg_schema
• Load a data model from Neo4j GraphRAG Python Package schema format
• Input:
• neo4j_graphrag_python_package_schema (dict): Neo4j GraphRAG Python Package schema dictionary
• Returns: DataModel object

📚 Example Data Model Tools

These tools provide access to pre-built example data models for common use cases and domains.

• list_example_data_models
• List all available example data models with descriptions
• Input: None
• Returns: Dictionary with example names, descriptions, node/relationship counts, and usage instructions

• get_example_data_model
• Get an example graph data model from the available templates
• Input:
• example_name (str): Name of the example to load ('patient_journey', 'supply_chain', 'software_dependency', 'oil_gas_monitoring', 'customer_360', 'fraud_aml', or 'health_insurance_fraud')
• Returns: ExampleDataModelResponse containing DataModel object and Mermaid visualization configuration

📝 Cypher Ingest Tools

These tools may be used to create Cypher ingest queries based on the data model. These queries may then be used by other MCP servers or applications to load data into Neo4j.

• get_constraints_cypher_queries
• Generate Cypher queries to create constraints (e.g., unique keys) for all nodes in the data model
• Input:
• data_model (DataModel): The data model to generate constraints for
• Returns: List of Cypher statements for constraints

• get_node_cypher_ingest_query
• Generate a Cypher query to ingest a list of node records into Neo4j
• Input:
• node (Node): The node definition (label, key property, properties)
• Returns: Parameterized Cypher query for bulk node ingestion (using $records)

• get_relationship_cypher_ingest_query
• Generate a Cypher query to ingest a list of relationship records into Neo4j
• Input:
• data_model (DataModel): The data model containing nodes and relationships
• relationship_type (str): The type of the relationship
• relationship_start_node_label (str): The label of the start node
• relationship_end_node_label (str): The label of the end node
• Returns: Parameterized Cypher query for bulk relationship ingestion (using $records)

💡 Prompts

• create_new_data_model
• Provide a structured parameterized prompt for generating a new graph data model
• Input:
• data_context (str): Description of the data and any specific details to focus on
• use_cases (str): List of use cases for the data model to address
• desired_nodes (str, optional): Node labels to include in the data model
• desired_relationships (str, optional): Relationship types to include in the data model
• Returns: Structured prompt that guides the agent through a 3-step process: analysis of sample data and examples, generation of the data model with validation, and presentation of results with visualization

🔧 Usage with Claude Desktop

💾 Released Package

Can be found on PyPi https://pypi.org/project/mcp-neo4j-data-modeling/

Add the server to your claude_desktop_config.json with the transport method specified:

"mcpServers": {

"neo4j-data-modeling": { "command": "uvx", "args": [ "mcp-neo4j-data-modeling@0.8.2", "--transport", "stdio" ] } }

🏷️ Namespacing Tools

The server supports namespacing the server tools:

"mcpServers": {

"neo4j-data-modeling-app1": { "command": "uvx", "args": [ "mcp-neo4j-data-modeling@0.8.2", "--transport", "stdio", "--namespace", "app1" ] }, "neo4j-data-modeling-app2": { "command": "uvx", "args": [ "mcp-neo4j-data-modeling@0.8.2", "--transport", "stdio", "--namespace", "app2" ] } }

With namespacing enabled:

• Tools get prefixed: app1-validate_node, app2-validate_node
• Each namespace operates independently

🌐 HTTP Transport Mode

The server supports HTTP transport for web-based deployments and microservices:

# Basic HTTP mode (defaults: host=127.0.0.1, port=8000, path=/mcp/)

mcp-neo4j-data-modeling --transport http

Custom HTTP configuration

mcp-neo4j-data-modeling --transport http --host 127.0.0.1 --port 8080 --path /api/mcp/

With namespace for multi-tenant deployment

mcp-neo4j-data-modeling --transport http --namespace myapp

Environment variables for HTTP configuration:

export MCP_TRANSPORT=http

export NEO4J_MCP_SERVER_HOST=127.0.0.1 export NEO4J_MCP_SERVER_PORT=8080 export NEO4J_MCP_SERVER_PATH=/api/mcp/ export NEO4J_NAMESPACE=myapp mcp-neo4j-data-modeling

🔄 Transport Modes

The server supports three transport modes:

• STDIO (default): Standard input/output for local tools and Claude Desktop
• SSE: Server-Sent Events for web-based deployments
• HTTP: Streamable HTTP for modern web deployments and microservices

🐳 Using with Docker

Here we use the Docker Hub hosted Data Modeling MCP server image with stdio transport for use with Claude Desktop.

Config details:

• -i: Interactive mode - keep