Claude in Microsoft Foundry is now generally available

Claude in Microsoft Foundry is the solution that businesses have been waiting for: real model choices, Azure-native management, easier procurement processes, and quicker route to value.

Most AI initiatives in large companies don’t hit a wall due to the quality of the models themselves. Instead, they tend to falter because of surrounding factors such as procurement issues, governance challenges, networking, and data management. Claude, now available on Azure through Microsoft Foundry, offers teams a quicker transition from testing to full-scale production.

Businesses can work with Claude using their current Azure account. This means they can rely on familiar authentication, billing, networking, governance, and data controls. Instead of worrying about the infrastructure, teams can concentrate on crafting applications that operate seamlessly using Claude in their existing environment.

This marks a significant leap forward for those looking to shift from AI trials to production-ready applications. Claude provides top-notch capabilities such as coding, agent work processes, and intricate reasoning. Together with Microsoft Foundry, businesses gain a robust framework for creating, assessing, deploying, and scaling these agents on Azure. This collaboration offers teams a reliable pathway to implement AI in production with superior model quality and the trusted Azure controls they are already accustomed to.

This latest announcement builds on the strategic partnership announced in November 2025 between Microsoft, NVIDIA, and Anthropic, aimed at broadening enterprise access to Claude utilizing NVIDIA’s accelerated computing. Claude operates on NVIDIA Blackwell Ultra systems, linked via InfiniBand networking, ensuring advanced AI infrastructure tailored for performance and efficiency.

How to Build with Claude Using Your Azure Account

Developers can engage with Claude via the Messages API, accessing key features like prompt caching, extended reasoning, and tool streaming. When building agents, the Foundry Agent Service utilises Claude as the central reasoning engine for executing complex tasks, multi-step planning, and tool integration within enterprise systems.

All inference takes place in Azure, with options available for both Global and US-specific data zones, catering to teams with specific data residency needs. Anthropic manages the inference process and acts as the data processor and SLA provider. The native availability of Claude through Foundry allows teams to engage in their usual Azure environment. They can authenticate easily with Microsoft Entra ID, implement Azure role-based access controls, manage permissions using existing governance policies, and monitor usage via familiar Azure management tools.

For highly sensitive tasks, zero data retention is also an option, ensuring that prompts and responses are not saved by Anthropic once the API call concludes. This also streamlines how Claude can be procured and utilised by commercial teams. Billing for Claude is done using Claude Consumption Units (CCU), presented as a single line item on your Azure invoice, while the MACC drawdown and per-model details in Foundry remain unchanged.

For many businesses, this combination of factors is just as important as model capability. The challenges in reaching production aren’t simply about whether a model is powerful enough but also whether teams can effectively procure, manage, secure, and scale it in their existing cloud environment. With Claude in Foundry, they unlock high-level capabilities while ensuring compliance, security, and governance aligned with enterprise standards.

Using Anthropic’s models on Azure has delivered the consistency and reliability that our enterprise customers need. The combination of leading model quality and enterprise-grade infrastructure makes Bolt a viable option for Fortune 500 companies.

—Gary Ballabio, Vice President, Partnerships, Bolt

Businesses Are Already Implementing Claude in Foundry

Companies aren’t just running isolated trials; they are establishing production systems and agents that require high throughput, dependability, governance, security, and scalability.

At NVIDIA, we deploy autonomous AI agents daily to expedite our operations and foster innovation. Anthropic’s Claude models offer strong reasoning, coding, and enterprise functionalities, which are crucial for complex technical tasks. With Claude now accessible in Microsoft Foundry on NVIDIA GB300 GPUs, more organisations can operate advanced, specialised AI agents meeting the performance, scale, and security demands required for production.

—Justin Boitano, Vice President and GM of Enterprise Computing, NVIDIA

Our customers express their requirements in straightforward English, and Momentic seamlessly verifies everything through our interface before a product release. We have found Claude’s Opus models particularly effective for this purpose, enabling us to handle millions of tokens per minute with the reliability our clients count on.

—Jeff An, Co-Founder and CEO, Momentic

Designed for Coding, Agents, and Complex Reasoning

Claude models are particularly adept at managing some of the fastest-growing AI needs in enterprises. For software developers, Claude excels at generating code, refactoring, debugging, crafting tests, and managing large-scale development processes. For teams devising agents, it facilitates multi-step reasoning, tool operation, planning, and executing tasks. Business teams can leverage it for thorough document analysis, research synthesis, and complex decision-making support.

Inside Microsoft Foundry, these features connect seamlessly with the wider Azure ecosystem. Using the Foundry Agent Service, teams can orchestrate goal-directed agents powered by Claude, enabling them to plan, call tools, and execute tasks across various enterprise systems. Innovations like the model router allow customers to direct queries efficiently to the most suitable Claude model, potentially saving up to 50% while enhancing user satisfaction. All this is underpinned by the Foundry Control Plane, which continuously evaluates agent responses to align with customer expectations, even blocking inappropriate responses before they reach users.

Collaborating with both Anthropic and Azure ensures we access unmatched capabilities and top-tier security. That’s exactly what the nuclear sector demands. We’ve managed to condense a safety analysis that typically required 200 human days into just a single day.

—Matt Huang, Founding Product Lead, Everstar

And with Microsoft IQ, agents can access real-time context from the enterprise, significantly elevating the value generated from each token. This helps enhance the impact that customers can achieve: tools like agent optimizer in Foundry Agent Service fine-tune the prompts that define agents for optimal performance, regardless of the underlying model.

Building a Solid Foundation for Enterprise AI on Azure

The next stage of enterprise AI will focus on production systems: coding agents, business process agents, research support, customer-oriented applications, and specialist workflows that operate reliably at scale. Achieving that requires more than just access to a model; it necessitates a comprehensive platform.

With the general availability of Claude in Microsoft Foundry hosted on Azure, customers have the opportunity to employ Anthropic’s cutting-edge models, orchestrating them as agents with Foundry Agent Service, supported by enterprise knowledge through Microsoft IQ, and leveraging the Azure controls, commitments, and infrastructure they already trust.