Think of it as a "bring your own workflow" starter kit for Foundry. It is not a new IDE and it is not a hosted service. It is a setup convenience, and a useful one if you have lost an afternoon to mismatched CLI versions.
What the Dev Pack installs
The Foundry Dev Pack is an all-in-one installer that prepares your machine for Foundry development. Microsoft's announcement lists four pieces:
- Foundry CLI (
azd ai): the terminal workflow for scaffolding, deploying, evaluating and automating agents. - Microsoft Foundry Skill: reusable Foundry workflows for compatible coding agents, including GitHub Copilot, Claude Code and Cursor.
- Foundry Toolkit for VS Code: installed only when VS Code is present.
- Foundry plugin for the GitHub Copilot app: installed only when that app is present.
Microsoft Learn spells out the boundary: DevPack doesn't install Visual Studio Code or GitHub Copilot App. Neowin also notes that Foundry Dev Pack doesn't download, install, and initialize a programming language. You still need Python or another runtime on the machine yourself.
The command-line tools in the pack include Azure CLI and Azure Developer CLI. Microsoft's own announcement describes Azure CLI (az), Azure Developer CLI (azd) and the Microsoft Foundry Extension for azd as part of that set.
Installing on Windows
On Windows the install is one line:
winget install Microsoft.FoundryDevPack
The other platforms use different commands:
- macOS:
brew install --cask microsoft/foundry/devpack && foundry-devpack install - Linux:
curl -fsSL [url]https://aka.ms/foundry-devpack-install.sh[/url] | bash
Microsoft suggests putting these commands in onboarding documents, dev containers and machine setup scripts. A third-party winget index lists the package as version 0.1.6 and covers winget upgrade --id Microsoft.FoundryDevPack --exact for updates. That index is not a Microsoft source, so check the version you get against Microsoft's documentation. A 0.x installer number alongside a "generally available" label is worth noticing, though Microsoft has not said how the installer version relates to the GA status of the tools it installs.
Learn also says Foundry DevPack is the recommended way to install the developer tools in one command, and you can also install each tool separately. Nothing is locked in.
Four ways to work
The announcement describes four entry points that share project files, so you can scaffold in one and debug or deploy from another.
Foundry Toolkit for VS Code
This is the visual, repository-centered option. Microsoft says you can:
- Browse Foundry project resources.
- Try models in the Model Playground.
- Scaffold an agent and press F5 to debug it.
- Inspect requests and responses in Agent Inspector.
- Use traces to find a problem, fix it and redeploy from the editor.
Models, agent code, tests and evaluations all stay in the repository.
Foundry CLI (azd ai)
This is the route for scripts, platform engineers and CI/CD. Agent configuration lives in azure.yaml. You can customize infrastructure with Bicep or Terraform when needed.
The announcement's example uses the basic Python Agent Framework sample. Its stated prerequisites are:
- Azure Developer CLI 1.37.0 or later
- A current
uvrelease - Python 3.13 or later
The sequence is:
azd auth loginazd ai agent initwith the sample template,--deploy-mode codeand--src ., run in an empty folderazd provision, after you confirm tenant, subscription, project, region, model and quotaazd ai agent run, thenazd ai agent invoke --local "<prompt>"from a second terminalazd deploy
Three more commands cover the rest of the loop. azd ai agent monitor --follow streams logs. azd ai agent doctor diagnoses setup and permission problems. azd pipeline config wires up GitHub Actions or Azure DevOps.
Foundry Canvas in the GitHub Copilot app
Canvas is a guided design surface where you choose the agent's model, tools, skills and guardrails. Copilot then generates project files you can inspect and edit.
Microsoft Learn labels Canvas as preview. Items marked preview are provided without a service-level agreement. Learn adds that Microsoft doesn't recommend preview features for production workloads. The Dev Pack is GA, but this piece of it carries a preview label in the documentation as of its November 5 update.
Learn also says Canvas needs the GitHub Copilot App with the microsoft-foundry plugin and a Foundry project. It supports hosted agents and prompt agents using the Microsoft Copilot harness.
Microsoft Foundry Skill
The skill gives coding agents Foundry-specific workflows so they lean on az, azd, MCP tools and your repository scripts. Microsoft says this helps agents avoid broad web searches and repeated trial and error, with fewer tokens and faster iteration. That is Microsoft's claim, and the announcement gives no measurements to back it up.
What still costs money
The Dev Pack installs tooling, but the work behind it still runs on Azure. Microsoft's CLI GA post from November 5 says you need an Azure subscription, Foundry permissions and model quota in your chosen region. It adds that provisioned resources and model usage incur Azure charges.
Two details matter in practice:
- "Local" is not offline. When you run the agent locally, model calls still use your Azure deployment, according to the CLI GA walkthrough. Local testing still hits a billable Azure deployment.
- Deploys happen remotely.
azd deployuploads your source to Foundry for a remote build, so no local container build is needed.azd deploy --previewlets you review hosted-agent configuration changes first.
The same walkthrough has advice for pipelines and cleanup:
- Grant the pipeline identity the required Azure permissions.
- Use a separate
azdenvironment for each stage. - Keep runtime credentials in pipeline secrets.
- Review the deletion before confirming
azd down. - Expect
azd downto delete the resource group and everything in it only if you created a new Foundry project for the demo. - Expect
azd downto leave the project and its resources in place, including the hosted agent, if you reused a project.
Analysis: who benefits
The main gain is consistency. A team can pin one install command in a setup script and get the same CLI, skill and editor integrations on every machine. Techzine reads it the same way: it describes the Dev Pack as primarily bundling existing Foundry and Azure tools into a single installation process. That is a fair reading. The announcement presents the pack as the recommended way to prepare for Foundry, not as a new capability.
The coding-agent angle also deserves some skepticism. The skill is guidance for an agent, not a guarantee that the plans or commands it produces are correct or safe. Microsoft itself says to review the agent's plans, file changes, commands and cost-bearing actions before they run. For anyone letting an agent provision Azure resources, treat that as a rule, not a suggestion.
Where to start
- Install Python 3.13 or later and Git, since the Dev Pack won't.
- Run
winget install Microsoft.FoundryDevPack. - Confirm your Azure subscription, Foundry permissions and model quota.
- Try the hosted-agent sample in an empty folder, and read
azure.yamlbefore runningazd provision. - Keep
azd ai agent doctorhandy if permissions or setup misbehave. - Run
azd downwhen the demo is over, after checking whether you created a new project or reused one.
Microsoft's Ignite sessions on Foundry Canvas and the Foundry Toolkit cover the same ground if you would rather watch it demonstrated.
References
- The Foundry Dev Pack is here: build Foundry agents your way Microsoft Foundry Blog · 2026-11-18T17:00:39+00:00
- Microsoft bundles AI development tools in Foundry Dev Pack - Techzine Global techzine.eu
- Microsoft is now offering Foundry Dev Pack for streamlined AI development - Neowin neowin.net