Microsoft Developer demonstrates how to use VS Code, GitHub Copilot, and the AI Toolkit to build, trace, and deploy AI agents to Azure with a workflow suitable for developers.

Building AI Agents with VS Code, GitHub Copilot, and Azure

In this session, you’ll learn to prototype, trace, and deploy an AI agent using the latest integrations available in Visual Studio Code. The workflow includes:

Workflow Overview

Starting with a prototype setup powered by GitHub Copilot.
Connecting Copilot to the AI Toolkit via an MCP server.
Using Copilot to:
- Suggest appropriate models
- Open and use the model playground
- Instrument code for detailed tracing
- Enable trace viewing to monitor agent actions
Deploying the agent to Azure and verifying end-to-end functionality

Step-by-Step Key Topics

VS Code & GitHub Copilot Integration
- Setting up Copilot extensions
- Using Copilot suggestions in conjunction with the AI Toolkit
Connecting to AI Toolkit via MCP
- Establishing a connection to the local MCP server
- Using the toolkit to manage and test models
Prototyping and Instrumentation
- Building the initial agent with Copilot-powered code generation
- Instrumenting the agent’s code for tracing
Trace Viewing
- Enabling and interacting with trace outputs to monitor agent behavior
Deployment to Azure
- Preparing deployment scripts and configuration
- Using Azure resources to host and operate the agent
Repeatable Workflow
- Steps for moving from local prototyping to full cloud deployment
- Recommendations on how to maintain and improve agents post-deployment

Hands-on Links and Resources

This workshop is ideal for developers looking to integrate AI capabilities into cloud-hosted agents using modern, developer-friendly tools while maintaining efficient traceability and repeatable deployment processes.