Microsoft Foundry Powers Agentic Era with General Availability of Production-Ready AI Agent Capabilities

Microsoft today announced the general availability of significant updates to Microsoft Foundry, its end-to-end platform for building, running, governing, and distributing AI agents. These advancements aim to streamline the transition of AI agents from experimentation to production, empowering organizations to leverage the full potential of generative AI within their existing workflows and infrastructure. The announcement, made in conjunction with Microsoft Build, underscores the company’s commitment to enabling developers to create, deploy, and manage AI agents seamlessly, without the need to piece together disparate tools and services.

More than 100,000 organizations are already utilizing Microsoft Foundry, with prominent companies like Adobe, Telefónica, and Tata Consultancy Services actively running agents in production environments. The platform’s evolution addresses a critical need in the burgeoning agentic era: the ability to integrate AI into real-world systems that are reliable, observable, and demonstrably aligned with business outcomes. Microsoft’s vision, first articulated at Microsoft Build, is to provide developers with a unified platform where they can build agents where they already work, run them on trusted infrastructure, and deploy them to end-users efficiently.

The latest set of updates, now generally available, deliver on this promise by consolidating frontier models, production agent runtime, enterprise-grade identity, security, and compliance controls, and distribution across Microsoft 365 into a single, cohesive platform. This integration is designed to eliminate the friction typically associated with stitching together disconnected tools and services, allowing organizations to accelerate their AI initiatives from concept to widespread adoption.

Why Foundry Stands Out as the Premier Agent Platform

Microsoft Foundry distinguishes itself as a comprehensive solution for the entire AI agent lifecycle. The platform is built upon three core pillars:

• Build: Enabling developers to create agents using their preferred frameworks and models, directly within their familiar development environments.
• Run: Providing a robust and scalable infrastructure for deploying and operating AI agents in production, with enterprise-grade security and compliance.
• Govern and Distribute: Offering tools for managing, optimizing, and distributing AI agents across an organization, ensuring responsible AI deployment and measurable business impact.

These pillars are interconnected in the latest Foundry updates, facilitating the seamless build, run, and scale of production-ready agents on a unified platform.

Empowering Developers: Build with Any Framework and Model

The agent development process is designed to commence within developers’ existing workflows, integrating directly with tools like GitHub Copilot and Microsoft Visual Studio (VS) Code. The Foundry Toolkit for VS Code, coupled with the Foundry skill, simplifies the deployment of agents to the Foundry platform. Whether teams opt for the Microsoft Agent Framework, the GitHub Copilot SDK, or the Claude Agent SDK, Foundry serves as the unified production destination.

A crucial aspect of agent capability lies in the underlying reasoning model. Microsoft Foundry offers access to a diverse range of industry-leading models, including frontier, open-source, and task-specific options, all accessible through a single platform. This allows teams to select the optimal model for each specific workload, ensuring both performance and cost-efficiency.

Introduction of GPT-5.6 Series and Enhanced Model Access

A significant development is the general availability of OpenAI’s GPT-5.6 series within Microsoft Foundry Models and Microsoft Foundry Agent Service. This latest iteration of the GPT model series introduces enhanced capabilities, providing organizations with greater flexibility to align model performance, cost, and specific business requirements. The GPT-5.6 series is available in three tiers:

• GPT-5.6 Sol: Designed for maximum capability, offering the most advanced reasoning and generative power for complex tasks.
• GPT-5.6 Terra: Strikes a balance between capability and cost, suitable for a broad range of demanding AI applications.
• GPT-5.6 Luna: Optimized for cost-effectiveness, providing strong performance for everyday tasks and high-volume usage.

This tiered approach allows organizations to move beyond a one-size-fits-all model deployment, matching the right model to the right job, thereby optimizing resource allocation and maximizing return on investment.

Microsoft’s commitment to delivering cutting-edge AI innovation where customers already operate is further demonstrated by the immediate availability of GPT-5.6 across various deployment options. The latest model is accessible through Global Standard and Global Priority Processing in all 28 global regions, Data Zones Standard, and Global Provisioned from day one. This ensures that customers can adopt the newest frontier AI advancements without needing to re-architect their existing deployments.

Pricing for GPT-5.6 Models

The pricing structure for the GPT-5.6 series in Microsoft Foundry is designed to offer transparency and predictability for deployment costs. The following table outlines the per-million-token pricing for input and output across the Sol, Terra, and Luna models within the Standard Global deployment:

Note: Output pricing is not detailed in the provided source snippet but is understood to be a critical component of token economics.

This tiered pricing model allows organizations to make informed decisions based on their specific use cases, balancing the need for advanced AI capabilities with budget considerations.

Global Reach and Data Sovereignty: The APAC Data Zone

Beyond offering a wider selection of models, Microsoft Foundry is expanding its operational footprint to accommodate diverse regulatory and performance requirements. The general availability of the Asia-Pacific (APAC) Data Zone for Microsoft Foundry marks a significant step in this direction. This new zone enables customers in the APAC region to run frontier OpenAI models while ensuring that data processing remains within designated Asia-Pacific regions. This eliminates the need for separate environments and accelerates the adoption of advanced AI without compromising data sovereignty or compliance mandates.

With Foundry now offering Global, Data Zone, and Regional deployment options, organizations can strategically align their AI adoption strategies with their specific sovereignty, compliance, performance, and scale requirements. This flexibility, combined with a consistent development and operations experience across all environments, provides a robust foundation for global AI deployments.

Hongsoo Kim, Chief Data and AI Officer (CDAO) at Viva Republica (Toss), commented on the significance of these advancements for financial institutions: "As financial institutions adopt AI, responsible data handling becomes foundational to trust. Microsoft Foundry’s APAC Data Zone allows us to keep data processing regionally anchored while accessing advanced AI models at scale. This gives us the confidence to accelerate AI innovation responsibly and reinforces our ambition to be a leading AI-powered financial platform in Asia."

Driving Impact with Action-Oriented Agents

A powerful model is merely the starting point for a production-ready AI agent. To translate AI capabilities into tangible business value, agents require a robust runtime environment, contextual understanding of business operations, governed access to tools, persistent memory across interactions, the ability to act on real-world events, and a clear path to end-users. Foundry integrates these essential components as built-in capabilities, designed to work in concert.

Key built-in capabilities within Foundry include:

• Action-Oriented Agents: Enabling agents to perform tasks and execute workflows within an organization’s systems.
• Context-Awareness: Allowing agents to understand and leverage relevant business context for more intelligent decision-making.
• Tool Integration: Providing secure and governed access to a wide array of tools and services that agents can utilize.
• Persistent Memory: Enabling agents to retain information across interactions, fostering continuity and personalized experiences.
• Event-Driven Operations: Allowing agents to respond to and act upon real-world events, facilitating dynamic automation.
• Seamless Distribution: Facilitating the deployment of agents to end-users through familiar Microsoft 365 applications and other channels.

Governing and Optimizing the AI Lifecycle with Enhanced Observability and Controls

The ability to see, improve, and secure an AI agent is paramount for successful production deployment. Microsoft Foundry prioritizes trust by embedding these capabilities directly into the platform, rather than placing the onus on individual developers. The latest release enhances the post-build lifecycle by providing comprehensive visibility into agent actions, enabling continuous improvement, and demonstrating their value.

New features for governing and optimizing AI agents include:

• Observability and Monitoring: Real-time insights into agent performance, usage patterns, and potential issues.
• Agent Debugging and Testing: Tools to diagnose and resolve problems, ensuring agent reliability.
• Performance Tuning: Capabilities to optimize agent responses, speed, and resource utilization.
• Cost Management and Optimization: Features to control and predict AI spending, ensuring efficient resource allocation.
• Responsible AI Tools: Mechanisms for ensuring fairness, transparency, and safety in AI deployments.

As agents scale from pilot programs to handling thousands of daily requests, Foundry equips teams with the necessary controls to manage costs predictably without leaving the platform. This cost management is underpinned by a focus on choice and optimization.

Foundry offers several mechanisms to optimize agent economics:

• Model Router: Intelligently matches each request to the most appropriate model, balancing capability and cost.
• Prompt Caching: Reduces redundant computation by storing and reusing results from identical prompts.
• PTU Spillover and Quota Optimization: Ensures service continuity during peak usage by intelligently managing provisioned throughput units (PTUs).
• Toolboxes: Limits the tools an agent accesses to only those required for a specific request, improving efficiency.
• Agent Optimizer: Fine-tunes prompts, skills, tools, and model choices against custom evaluators to enhance performance and reduce costs.

Beyond cost efficiency, Foundry provides a clear view of an agent’s return on investment (ROI). The ROI dashboard connects business value, usage metrics, and operational costs, enabling teams to assess whether a production agent is generating more value than it consumes in resources and identify areas for further optimization.

For a deeper understanding of token economics and agent optimization, a new Microsoft Mechanics episode on the subject is available.

Real-World Impact: Organizations Building on Foundry

The adoption of Microsoft Foundry is not limited to experimentation; it is actively driving tangible business outcomes for a wide range of organizations, from digital-native startups to global enterprises. Companies like Adobe, Telefónica, and Tata Consultancy Services are among those leveraging Foundry to bring AI agents into production.

The common thread among these organizations is a significant acceleration in their AI deployment timelines. What once took weeks for integration, security, and deployment can now be accomplished in days. This rapid deployment is facilitated by an infrastructure that meets stringent compliance standards and enables agents to reach users through the tools they already trust and use daily.

Getting Started with Microsoft Foundry

All the features and capabilities discussed are now live and accessible within Microsoft Foundry.

Developers can begin their journey by consulting the comprehensive documentation and Microsoft Learn courses available. The Quickstart guide provides a hands-on walkthrough for setting up, testing, and deploying a production-ready hosted agent end-to-end, allowing developers to get started in minutes.

For a structured learning experience, the "AI Agents for Beginners" curriculum offers a 12-lesson program. Further practical learning can be achieved through guided labs such as "Develop AI Agents in Azure," "Hosted Agents Workshop (.NET)," the "Foundry Toolkit for VS Code and hosted agents workshop," and the "ZavaShop Supply Chain Workshop." To ensure the quality and reliability of deployed agents, a practical guide on "Evaluating AI Agents with Microsoft Foundry" is also available.

For those seeking a visual explanation of how to operationalize AI agents from deployment to real-world impact, Jeff Hollan’s Microsoft Mechanics episode, "Foundry Agent Service + Microsoft Agent Framework Explained," offers valuable insights.