Microsoft Ignite 2026 wraps up today, and the announcement I keep coming back to isn't a keynote demo — it's a setup tool. On November 18, Microsoft made the Foundry Dev Pack generally available: one install that puts the Microsoft Foundry developer tools on your machine.
If you've been following this week, you know the pattern: Tuesday I shared the free Learn path for building agents with Foundry. The Dev Pack is the step before that path — the part that usually eats an afternoon of installs and version mismatches.
What one install actually gives you
The Dev Pack bundles four things, and they share the same project files, so you can move between them without starting over:
-
Foundry CLI (
azd ai) — the terminal workflow: scaffold, run locally, deploy, monitor, all from repeatable commands stored inazure.yaml. - Foundry Toolkit for VS Code — a visual workflow with a Model Playground, Agent Inspector, and F5 debugging, installed when VS Code is present.
- Foundry plugin for the GitHub Copilot app — Foundry Canvas, a guided design surface where you pick the model, tools, skills and guardrails and Copilot generates the project files.
- Microsoft Foundry Skill — reusable workflows for coding agents like GitHub Copilot, Claude Code and Cursor, so your agent helper already knows Foundry's commands.
Two honest boundaries from Microsoft's own announcement: the pack doesn't install VS Code or the GitHub Copilot app themselves, and it doesn't install a programming language — you still need Python (3.13 or later for the samples) on your machine.
Install it
Windows (PowerShell or terminal):
winget install Microsoft.FoundryDevPack
macOS:
brew install --cask microsoft/foundry/devpack && foundry-devpack install
Linux:
curl -fsSL https://aka.ms/foundry-devpack-install.sh | bash
That's the whole setup. One command, then you're building instead of configuring.
What you need before you deploy
The tools are free to install, but the agent itself runs on Azure, so be clear-eyed about the rest: you need an Azure subscription (students can start with the free Azure for Students offer), a Foundry project with a deployed model, and model quota in your region. Provisioning and model usage incur Azure charges — start local, deploy when you're ready, and check the cost in azure.yaml before you provision.
The inner loop: scaffold to deployed agent
Microsoft's own walkthrough uses the basic Python Agent Framework sample. From an empty folder:
azd auth login
azd ai agent init --template https://github.com/microsoft-foundry/foundry-samples/blob/main/samples/python/hosted-agents/agent-framework/responses/01-basic/azure.yaml --deploy-mode code --src .
azd provision
azd ai agent run
azd ai agent invoke --local "Write a haiku about deploying cloud applications."
azd deploy
What each step does: init scaffolds the sample and writes the project config; provision creates the Azure resources declared in azure.yaml (review that file first — it declares the model deployment and resources you're about to create); run starts the agent locally and opens the Agent Inspector in your browser so you can chat with it; invoke --local sends it a prompt from the terminal; deploy packages your code and ships it to Foundry Agent Service as a hosted agent. After deploying, azd ai agent invoke (without --local) talks to the hosted version, azd ai agent monitor --follow streams its logs, and azd ai agent doctor diagnoses setup and permission problems.
Pick your entry point, keep the same project
This is the part I like most: the tools share project files. Scaffold the sample in the terminal, open the same folder in VS Code and debug it with F5 and the Agent Inspector, or hand the same project to your coding agent with the Foundry Skill and ask it to add a tool. Nothing to re-create, nothing to reconfigure.
If you're not sure where to start, start in the terminal — the commands above are the shortest path to a running agent. If you prefer a visual flow, the VS Code Toolkit is the same project with buttons. And if you want to design before you code, Foundry Canvas in the Copilot app is the guided route.
Go further
Once the sample is running, the next step is the hosted-agent quickstart on Microsoft Learn — it walks the same flow in full detail, including the Python SDK and .NET paths: https://learn.microsoft.com/azure/foundry/agents/quickstarts/quickstart-hosted-agent?pivots=azd&wt.mc_id=studentamb_616890
And if you want the concepts behind the commands, the free Learn path on developing AI agents with Foundry is still the best next read: https://learn.microsoft.com/training/paths/develop-ai-agents-azure/?wt.mc_id=studentamb_616890
Setup was the boring part of building agents. As of this week, it's one command. The interesting part — what your agent actually does — is now the only hard bit left.
What would you build first?
Top comments (0)