Asha Sharma introduces the integration of Anthropic’s Claude models into Microsoft Foundry on Azure, offering developers extensive AI options and practical tools for building enterprise-ready intelligent agents.

Introducing Anthropic’s Claude Models in Microsoft Foundry: Expanding AI on Azure

Author: Asha Sharma

Microsoft has expanded its AI offerings on Azure by partnering with Anthropic to deliver Claude models — alongside OpenAI’s GPT models — through Microsoft Foundry. Azure is now the only cloud provider offering both Claude and GPT frontier models on a single platform, greatly increasing flexibility and capability for developers and organizations building next-generation AI solutions.

Why This Matters

  • Model Choice: Developers can select from a variety of advanced models to match each use case, from chatbots to research agents and enterprise workflows.
  • Unified Platform: Foundry enables the use of any model, framework, and enterprise control within a governed environment.
  • Enterprise-Ready: Models are operated with strict governance, observability, and security, ensuring solutions are deployable at scale.

Overview of Anthropic’s Claude Models

Model Strengths Ideal Use Cases
Claude Haiku 4.5 Fast, cost-efficient Free tier user experiences, real-time, coding/research/financial agents
Claude Sonnet 4.5 Complex agent reasoning, coding proficiency Long-running agents, coding, cybersecurity, financial analysis, research
Claude Opus 4.1 Specialized, advanced problem-solving Advanced coding, long-horizon/complex tasks, AI agent development

Key features:

  • All models use Constitutional AI for safety and can be managed through Foundry’s enterprise-grade governance.
  • Suitable for trustworthy automation in enterprise workflows, from customer support to coding agents.

Evolving to Intelligent Agents

The adoption of AI is rapidly shifting from static apps to flexible, agentic systems. Microsoft Foundry bridges the operational gap, letting organizations deploy, monitor, and manage powerful AI agents within secured and governed workflows on Azure.

“Having Claude’s advanced reasoning alongside GPT models on one platform gives us flexibility to build scalable, enterprise-grade workflows…” — Michele Catasta, Replit

Claude in Foundry Agent Service

Developers can utilize Claude models within Foundry Agent Service for:

  • Multi-step workflow planning
  • Integration with productivity tools via Model Context Protocol (MCP)
  • Automated data operations (classification, summarization, extraction)
  • Dynamic model selection with model router
  • Governance and oversight for agent fleets

Additionally, Claude Code enables developers to use Claude as an AI-driven coding agent for complex development workflows and automation.

Example Use Case

If a deployment fails, Claude (via Foundry) can:

  • Query Azure DevOps logs
  • Diagnose the failure
  • Recommend a fix
  • Trigger patch deployment, all within a governed Azure environment

Claude Skills: Modular, Composable AI Abilities

With the Claude API, developers create Skills — reusable, modular components that:

  • Include instructions, optional Python/Bash code, and linked assets/APIs
  • Automate tasks such as report generation, dataset cleaning, or PowerPoint summarization
  • Are version-controlled and governed within Foundry for enterprise reliability

Advanced Context with Deep Research

Deep Research extends Claude’s reasoning to live and internal enterprise data, producing grounded, nuanced analysis and automating analytical workflows.

Pricing

Model Input/1M Tokens Output/1M Tokens
Claude Haiku 4.5 $1.00 $5.00
Claude Sonnet 4.5 $3.00 $15.00
Claude Opus 4.1 $15.00 $75.00

All models available in East US 2 and West US Azure regions.

Getting Started

Conclusion

Microsoft’s partnership with Anthropic makes Azure a leader in providing powerful, flexible AI models for enterprise and developer use. Foundry ensures these capabilities are secure, governed, and ready for demanding real-world applications.

This post appeared first on “The Azure Blog”. Read the entire article here