Table of Contents

Getting started

This article walks you through creating a semantic model in Flow and making it available to AI agents.

1. Create a semantic model

  1. In the Portal, open your workspace and select Semantic models. This opens the semantic model manager in Flow.

    Navigating to semantic models from the Portal

  2. Create a new semantic model by entering a name and selecting a connection type. SQL Server is currently the only supported connection type. Choose an existing connection or create a new one.

    Creating a new semantic model

2. Add a dataset and metrics

In the explorer on the right, add a new dataset. Then configure it by adding columns and metrics.

Adding a dataset to the semantic model

3. Add dimensions (optional)

Add dimensions to model a star schema. Dimensions provide the descriptive attributes (the who, what, when, where and how) that give context to the facts in your datasets. Each dimension has a primary key that defines its relationship to the dataset (fact) table.

4. Add derived metrics (optional)

Add derived metrics to calculate new values from dataset metrics using formulas.

Adding a derived metric

5. Connect an MCP client

Connect an MCP client, such as Claude or Microsoft 365 Copilot, to query the semantic model using natural language.