Foundry Friday AMA · Jun 27, 2025 · Model Context Protocol

[nitya]

⁉️ Have questions on Model Context Protocol (MCP)? Add them here! We'll post a recap after the AMA on this thread

---

AMA on Model Context Protocol (MCP)

This is part of the #ModelMondays series where we put the spotlight on a new model-related topic each week.

🌟🌟 See #54 for the full Foundry Fridays AMA schedule 🌟🌟

---

Event Details

Curious about how to extend your AI models and agents with powerful new capabilities? This session dives into the Model Context Protocol (MCP), an open standard for connecting AI models to external tools and data. Den Delimarsky will walk you through the MCP specification, security best practices, and how MCP is used within Microsoft Azure to build agentic applications. Learn how to consume existing MCP servers, develop your own, and unlock new integration patterns for intelligent apps.

• 1️⃣ | Register for the Friday AMA - 1:30pm ET
• 2️⃣ | Watch the Monday Livestream - 1:30pm ET
• 3️⃣ | Learn more About Model Mondays - Season 1 Recaps + Season 2 Schedule

Related Resources

• Build Agents using Model Context Protocol on Azure
• How to use the Model Context Protocol (MCP) tool
• Get started with .NET AI and the Model Context Protocol

---

[nitya]

Catch Up On Monday Livestream

1. Watch: Model Mondays Playlist
2. Download: Speakerdeck Slides
3. Read: dev.to Blog
4. Join: Foundry Fridays AMA 👈🏽👈🏽👈🏽

[Yeok-c]

Hey was just trying out the mcp as a tool as demonstrated here:
https://gitmcp.io/Azure/azure-rest-api-specs

I'm getting the errors

(UserError) The tool 'mcp' is not supported it must be one of the following: code_interpreter, file_search, function, bing_grounding, azure_ai_search, openapi, connected_agent, azure_function

Is it because this feature is only available in the newest or a nightly build?

[nitya]

@Yeok-c I have not used that tool but have you tried out the Official GitHub MCP server to work with the GitHub repos instead?

For the specific error you identified - can you tell me the steps you took so I can see if I can reproduce the issue? Are you using this with a specific client or IDE? I typically use VS Code with Agent Mode to explore ideas but was not sure what your client setup was.

DO let me know and would love to follow it up.

[nitya]

S2:E02 – Model Context Protocol (Livestream Transcript Summary)

This post was generated with AI help and human revision & review. To learn more about our motivation and workflows, please refer to this document in our website

Click Banner To Watch Replay

Abstract

This episode of Model Mondays explores the Model Context Protocol (MCP), a foundational standard for connecting AI models to external data and tools. Den Delimarsky provides a deep dive into MCP, focusing on secure server implementation, the new authorization specification, and best practices for protecting AI infrastructure in the cloud.

News Highlights

• Cohere models are now available on managed compute in Foundry, enabling secure, scalable deployments on Azure infrastructure.
• Prompt Shields and Azure AI Content Safety provide comprehensive protection against prompt injection and offensive outputs.
• OpenAI’s 03 Pro model and Codex Mini are now available, offering improved performance and specialized code generation.
• MCP has been upgraded with model-based access control, better inference routing, and enhanced telemetry for observability and governance.
• The new MCP authorization spec separates resource and authorization servers, simplifying secure deployments and leveraging standard OAuth 2.1 flows.

Tech Spotlight

Den Delimarsky explains the core purpose of MCP: providing a universal protocol for supplying context to language models, making it easy to connect LLMs to diverse data sources. He details the importance of secure MCP server deployments, especially in remote/cloud scenarios, and the challenges of managing user context and authorization.

The new MCP authorization specification enables developers to use existing identity providers (like Entra ID, Okta, AWS Cognito) for token issuance and validation, reducing the need for custom security logic. Den demonstrates the flow of client-server interactions, the structure of protected resource metadata (PRM) documents, and the benefits of off-the-shelf security solutions. The session emphasizes developer ergonomics, security best practices, and the evolving MCP ecosystem.

Takeaways

• MCP standardizes how LLMs access external data and tools, enabling scalable, secure AI applications.
• The new authorization spec allows developers to use existing identity providers, reducing custom security work.
• Secure MCP deployments require proper token validation, audit logging, and adherence to best practices.
• Local MCP servers are simpler but less relevant for cloud-scale, multi-user scenarios.
• The MCP ecosystem is rapidly evolving, with new specs, SDKs, and best practices being adopted across tools and platforms.

Related Resources

• Security best practices for MCP
• Understanding and mitigating security risks in MCP implementations
• Connect to Model Context Protocol (MCP) Servers (Preview)
• What is the Azure MCP Server (Preview)?
• Register and discover remote MCP servers in your API inventory

[nitya]

Friday AMA Recap: Model Context Protocol (MCP)

This post was generated with AI help and human revision & review. To learn more about our motivation and workflows, please refer to this document in our website

Attendees

• Nitya Narasimhan, PhD [Microsoft]
• Sharmila Chockalingam [Microsoft]
• Lee Stott [Microsoft]
• Nick Brady [Microsoft]
• Community members from various locations

Locations

• New York, USA
• The Netherlands (originally Romania)
• Vancouver, Canada
• San Francisco, USA
• Ukraine (living in Poland)
• Kelowna, Canada
• Uzbekistan
• Philly, USA
• Morocco
• UK
• India

Core Topics Discussed

• Introduction to Model Context Protocol (MCP)
• Security and authorization in MCP
• Use cases for MCP in Microsoft Azure
• Community experiences with MCP (deployment, usage, integration)
• MCP in Playwright for UAT and custom server deployments

Key Takeaways

1. MCP is an open protocol for integrating tools and context into LLM workflows.
2. Security and authorization are critical for MCP server deployments.
3. MCP is being used in various scenarios, including Playwright for UAT and custom servers.
4. Community members are actively deploying and experimenting with MCP servers.
5. Microsoft Azure provides resources and templates for building and consuming MCP servers.

Key Resources

• Replay of Den's Talk
• Discussion Forum & Recap
• Azure MCP Server
• OpenAI MCP Agent Building Block AI template
• MCP Container App Building Block AI template

Microsoft Resources

• Build Agents using Model Context Protocol on Azure
• How to use the Model Context Protocol (MCP) tool
• Connect to Model Context Protocol (MCP) Servers
• Get started with .NET AI and the Model Context Protocol