Develop Microsoft Fabric notebooks with the Fabric Notebook custom agent in Visual Studio Code

Note

Fabric Notebook custom agent in Visual Studio Code is currently in preview.

The Fabric Notebook custom agent is a specialized agent that you can select in the GitHub Copilot Chat experience in Visual Studio Code. It helps you author Microsoft Fabric notebooks with suggestions and code generation that are aware of Fabric notebook context.

Compared to a general-purpose coding agent, the Fabric Notebook custom agent understands Fabric notebook patterns. For example, it recognizes the built-in spark variable that represents your current Spark session, so it can suggest code that uses the existing session instead of creating a new one.

It also helps with common Fabric data access patterns, such as using relative paths for the default Lakehouse and full ABFSS paths for nondefault Lakehouses.

When to use this agent

Use the Fabric Notebook custom agent when you need AI assistance for notebook authoring tasks, such as generating Spark code, refining notebook logic, and troubleshooting notebook code with Fabric-specific context.

Use more Fabric Data Engineering VS Code extension features when you need workspace and artifact operations, such as browsing Fabric items, opening notebooks, and managing resources. To learn what the extension is and what it supports, see What is the Fabric Data Engineering VS Code extension?.

Prerequisites

• Install the Fabric Data Engineering VS Code extension.

Select the Fabric Notebook custom agent

Follow these steps in order to select the session type, agent, and model in GitHub Copilot Chat.

1. Make sure the Fabric Data Engineering VS Code extension is installed.

   This extension is required for the Fabric Notebook custom agent to appear.

2. Open a Fabric notebook in Visual Studio Code.

3. Open GitHub Copilot Chat.

4. In the session type selector, choose Local.

   The FabricNotebook custom agent supports only the Local session type. Background and Cloud aren't supported for this agent.

5. In the agent selector, choose FabricNotebook.

   

6. In the model picker, choose one of the supported base models: Claude Sonnet 4.5, Claude Opus 4.6, or GPT-5.2.

   Important

   Use one of these base models with the Fabric Notebook custom agent: Claude Sonnet 4.5, Claude Opus 4.6, or GPT-5.2. If another model is selected, switch to one of these supported models.

Use the Fabric Notebook custom agent

After you select FabricNotebook, you can start using it immediately in GitHub Copilot Chat by entering your own prompts. You don't need to use sample prompts to work with the agent.

The sample prompts and saved prompts are optional helpers in the FABRIC DATA ENGINEERING extension view.

1. In the Activity Bar, select the FABRIC DATA ENGINEERING extension view.

2. In that panel, expand AGENT PROMPTS - FABRIC DATA ENGINEERING to view sample prompts.

   Sample prompts help you start common tasks quickly. For example, you can use prompts for exploring and validating data, cleaning and preparing data, and similar notebook workflows.

   

3. To save prompts for reuse, under AGENT PROMPTS - FABRIC DATA ENGINEERING, expand Saved Prompts.

   Use this section to store your frequently used prompts and access them quickly later.

Related content

• What is the Fabric Data Engineering VS Code extension?
• Develop, execute, and manage Microsoft Fabric notebooks
• Create and manage Microsoft Fabric notebooks in Visual Studio Code