Build and Ship a Hosted Agent with VS Code AI Toolkit & Microsoft Foundry

Microsoft Developer demonstrates an end-to-end workflow for building an agentic AI solution with the VS Code AI Toolkit and Microsoft (Azure AI) Foundry—from model selection and prototyping to hosting, deployment, evaluation, red teaming, and monitoring.

Overview

This video walks through an end-to-end (E2E) workflow for an AI engineer to build, extend, deploy, and operationalize a hosted agent using the VS Code AI Toolkit and Microsoft Foundry (Azure AI Foundry).

Step-by-step flow covered

1) Model selection

• Use GitHub Copilot and a Model Catalog to select a model.

2) Prototype an agent in the UI

• Prototype an agent in the Agent Builder UI
• Add tools to the agent
• Test prompts in a playground
• Ground the agent on business data

3) Move from UI prototype to code (hosted agent template)

• Migrate the UI-based agent to a hosted agent template
• Extend the agent with custom Python functions
• Debug locally

4) Deploy and operationalize in Foundry

• Deploy to Foundry
• Run evaluation runs
• Perform AI red teaming to assess attacks
• Use monitoring dashboards to track:
  • Quality
  • Usage
  • Latency
  • Cost
  • Fleet-wide agent behavior (across multiple agents)

Links

• Learn more: https://aka.ms/zava-sales-manager-hos...
• AI Toolkit: https://aka.ms/aitoolkit
• Microsoft Foundry docs: https://aka.ms/microsoft-foundry-doc

Event metadata

• eventID: 26884
• Hashtags: #microsoftreactor #learnconnectbuild