Click on "Deploy Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@Fabric Ontology MCP Server Discover workspace data for my Sales workspace"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
Production-ready MCP server for full CRUD control of Ontology items in Microsoft Fabric.
Python ≥ 3.11
Azure CLI (az) installed and logged in (az login)
Access to a Microsoft Fabric workspace with Ontology items
Related MCP server: Microsoft Fabric MCP Server
cd "Ontology MCP Server"
pip install -e .# Run via entry point
fabric-ontology-mcp
# Or directly
python -m src{
"servers": {
"fabric-ontology": {
"command": "python",
"args": ["-m", "src"],
"cwd": "/path/to/Ontology MCP Server"
}
}
}Tool | Description |
| List all Fabric workspaces accessible to you |
| List items in a workspace (filter by type: Eventhouse, Lakehouse, etc.) |
Tool | Description |
| List ontologies in a workspace |
| Get ontology metadata |
| Create a new ontology |
| Update display name / description |
| Delete an ontology (soft or hard) |
| Get full decoded definition (entities, relationships, bindings) |
| Replace entire definition from JSON |
Tool | Description |
| List all entity types in an ontology |
| Get a single entity type with its bindings, documents, overviews, links |
| Create a new entity type with properties |
| Rename an entity type or change its display name property / ID parts |
| Delete an entity type (and its relationships) |
Tool | Description |
| Add a property to an entity type |
| Rename a property or change its value type |
| Remove a property from an entity type |
Tool | Description |
| List relationships in an ontology |
| Get a single relationship with its contextualizations |
| Create a relationship between entity types (validates both exist) |
| Rename a relationship type |
| Delete a relationship type |
Tool | Description |
| List bindings for an entity type |
| Bind an entity type to a Lakehouse or Eventhouse table |
| Remove a data binding |
Tool | Description |
| Attach a document URL to an entity type |
| List all documents attached to an entity type |
| Remove a document by URL |
| Get the current overview configuration |
| Configure overview widgets for an entity type |
| Get the current resource links |
| Link Power BI reports to an entity type |
Tool | Description |
| List contextualizations for a relationship type |
| Define how a relationship is materialized from data |
| Remove a contextualization |
Tool | Description |
| Get cluster URI, database name, and parent Eventhouse |
| List tables in a KQL database |
| Get column names and types for a table |
Tool | Description |
| List all tables in a Lakehouse (via OneLake Table API) |
| Get columns with types and Ontology valueType mapping |
| Full scan — discover all Lakehouses + Eventhouses, all tables and schemas |
Ontology planning workflow: Call
discover_workspace_datato scan a workspace, then use the returned schemas to design entity types, properties, relationships, and data bindings using the CRUD tools above.
Tool | Description |
| Preview first N rows from an Eventhouse table |
| Row count, distinct counts, null rates, sample values, min/max per column |
| Preview first N rows from a Lakehouse table (Spark SQL via Livy) |
| Row count, distinct counts, null rates, sample values, min/max per column |
Note: First Lakehouse profiling call takes ~30-60s for Spark session startup. Subsequent queries reuse the session and are fast.
src/
├── __init__.py
├── __main__.py # python -m src entry point
├── auth.py # Azure CLI token acquisition (per-resource caching)
├── definition_utils.py # Base64 encode/decode for ontology definition parts
├── fabric_client.py # Async Fabric REST API client
├── kusto_client.py # Async Kusto REST query client
└── server.py # MCP server with all toolsAll tools validate inputs before calling the Fabric API:
Names must match ^[a-zA-Z][a-zA-Z0-9_-]{0,127}$ (entity types, properties, relationships)
Value types must be one of: String, Boolean, DateTime, Object, BigInt, Double
JSON parameters return clear error messages on parse failure
Entity existence is checked when creating relationships (both source and target must exist)
Uses Azure CLI tokens. Make sure you're logged in:
az loginTokens are cached per resource (Fabric API and Kusto clusters are separate audiences) and automatically refreshed when near expiry.
MIT
This server cannot be deployed
List datasets, schemas, run APL queries, and use prompts for exploration, anomalies, and monitoring.
Governed BI MCP. Ask questions of live company data and list workspace sources via OAuth.
Your Databricks Lakehouse in natural language: run SQL on your SQL warehouses, track long-running qu
Data-ontology maps of your business systems, served to AI agents over MCP.
