Edit

Share via

New Microsoft Foundry portal general availability overview

The new Microsoft Foundry portal is now generally available (GA). This milestone marks a shift from pilot-focused usage to secure, reliable, enterprise-ready production usage for core scenarios.

Foundry is designed for teams that need to build, deploy, and operate AI systems at scale, with governance, security, and operational controls integrated throughout the lifecycle. Foundry unifies the end-to-end lifecycle across Discover, Build, and Operate so teams can move faster without trading off reliability, compliance, or operational rigor.

Prerequisites

Before you standardize on GA features for production, make sure you:

  • Understand your required scenarios across model deployment, agent development, and operations.
  • Identify any current dependencies on preview-only or classic portal experiences.
  • Define your organization policy for using only GA capabilities in production.
  • Review migration guidance for existing Azure OpenAI and Foundry (classic) portal workloads.
  • Confirm required role assignments for your teams and service identities. For role details, see Role-based access control for Microsoft Foundry.
  • Define how your organization restricts preview feature access in production environments. For guidance, see Disable preview features in Microsoft Foundry.

Key terms used in this article

  • GA: Generally available features supported for production use.
  • Preview: Features that are not yet generally available.
  • Foundry projects: Workspace containers that organize your AI assets, deployments, and agent configurations within the new Foundry experience.
  • AOAI: Azure OpenAI resources and workflows.

What GA means for customers

At GA, the new Microsoft Foundry portal provides:

  • Production-ready core platform with validated end-to-end core scenarios.
  • Enterprise capabilities such as RBAC, audit logs, compliance controls, monitoring, alerting, virtual network integration. Also API keys are supported for all areas except for evals, dataset tab, content understanding, agents and workflows.
  • Governed lifecycle consistency across the portal, APIs, SDKs, CLI, and developer tools.
  • Defined GA scope for Foundry projects, with out-of-scope capabilities continuing in Foundry (classic) portal.

For governance-sensitive production workloads, use Microsoft Entra ID with RBAC for role-based access control. API key-based access is available, but it doesn't provide the same role-based permission granularity. For billing and cost management details, see Plan and manage costs for Microsoft Foundry.

GA scope by project type

At GA, the new Foundry experience supports Foundry projects for core end-to-end scenarios. Confirm that your target regions support the models and features you need. For region details, see Feature availability across cloud regions.

For scenarios not yet available in the new Foundry portal, you can continue to use Foundry (classic) portal to maintain continuity while capabilities continue to evolve.

Core scenarios at GA

Core GA coverage includes:

  • Model core flows: Discover models, deploy models, run inference, manage deployments, and transition to agent-based workflows.
  • Agent development: Build agents and integrate evaluations, tracing, monitoring, red teaming, and fine-tuning where supported.
  • Operate experiences: Manage agents and assets, enforce policies, and manage quota and administration features where supported.

Feature readiness at GA

Important

The status values in this section include roadmap-sensitive items. Confirm current status before making production commitments.

The following table summarizes feature readiness. Most core capabilities across Home, Discover, Build, and Operate are GA, while some capabilities remain in Preview.

Area Feature Status
Home All GA
Discover Overview GA
Discover Model GA
Discover Tools GA
Discover Solution Templates GA
Discover Agent Manifests Preview
Discover Search GA
Discover Ask AI Preview
Build Agents GA (minus Voice Live, traces in agent builder in Preview)
Build Workflows Preview
Build Models GA
Build Tracing and tracing VNet Preview
Build Optimization (cluster analysis) Preview
Build Fine-tuning GA
Build Tools GA (check label on individual tools in the catalog to determine if they are GA or Preview)
Build Knowledge Preview
Build Data GA (minus stored completions in Preview)
Build Evaluations GA
Build Memory Preview
Build Guardrails Agents = Preview; Models = GA; Controls and intervention = Preview
Build Monitoring Preview
Build Red teaming GA
Build AI services speech playgrounds GA
Operate Overview Preview
Operate Assets Preview
Operate Compliance Preview
Operate Quota GA
Operate Admin GA (minus AI Gateway in Preview)
Docs All GA

Unsupported at GA

The following items are out of scope at GA for the new Foundry portal and require the classic portal:

  • Standalone Azure OpenAI or other single-service resources that aren't connected to a Foundry project.
  • Assistant creation and authoring in the new Foundry experience.
  • Listing AOAI evaluation files as datasets for upgrade workflows.
  • Audio playground.
  • AI service fine-tuning.
  • Content Understanding.
  • Prebuilt prompts in video playground.
  • Adding data directly from the Data tab (users can add data during agent creation workflows).
  • Private/Government cloud support for the new Foundry experience.

FAQ

What does general availability mean for Microsoft Foundry?

GA means the new Foundry portal is supported for production use for defined core scenarios in Foundry projects, with validated end-to-end experiences, enterprise support readiness, and operational reliability.

Which projects are supported at GA?

At GA, the new Foundry experience supports Foundry projects with end-to-end coverage for core scenarios. Other resource types can continue in the Foundry (classic) portal where needed.

Are all Foundry features GA?

No. GA covers validated core experiences and required enterprise features. Some capabilities remain in public preview.

How do I disable preview features?

Use your organization controls to limit production environments to general availability supported capabilities, and validate current feature status before rollout decisions. For guidance on hiding preview features with tags or blocking them with custom RBAC roles, see Disable preview features in Microsoft Foundry.

What is the experience for existing Azure OpenAI users?

If you have existing Azure OpenAI resources, you can continue to use classic portal for unsupported workflows while you plan your upgrade to Foundry projects.

For upgrade guidance, see Upgrade Azure OpenAI to Microsoft Foundry.

For project migration guidance, see Migrate from hub-based to Foundry projects.

Are assistants supported in Foundry projects?

Agents v2 are supported in the new Foundry UI. Existing assistants and v1 agents aren't supported in the new Foundry experience. To use or edit assistants, continue using Foundry (classic) portal until assistant upgrade is available.

Can customers use Foundry GA through APIs and developer tools?

Yes. Foundry provides support across portal, APIs, SDKs, and CLI for GA-supported scenarios.

To get started, see Microsoft Foundry SDKs and Microsoft Foundry API.

Is GA the final state of Microsoft Foundry?

No. GA is a production milestone, not an endpoint. Microsoft continues to expand workflow authoring, operations, and governance capabilities based on customer feedback and production usage.

Validate GA-only usage

Before production rollout, validate the following:

  • Required scenarios in your workload map to capabilities marked GA in this article.
  • Dependencies on Preview features are documented and approved for nonproduction use only.
  • Role assignments and authentication model are aligned to your governance policy, especially where API keys are used.
  • Target-region model and feature availability are confirmed in Feature availability across cloud regions.
  • Teams supporting migration scenarios have a documented path between the new Foundry experience and Foundry (classic) portal workflows.

Common rollout pitfalls

Next steps

  • Last updated on