Edit

Share via


Azure MCP Server tools for Azure Cloud Architect

The Azure Model Context Protocol (MCP) Server lets you design cloud architectures through guided requirements gathering and receive optimal Azure solution recommendations with natural language prompts.

Azure Cloud Architect helps you design scalable, resilient Azure solutions and apply guidance from the Azure Architecture Center; for more information, see Azure Architecture Center documentation.

Note

Tool parameters: The Azure MCP Server tools define parameters for data they need to complete tasks. Some of these parameters are specific to each tool and are documented below. Other parameters are global and shared by all tools. For more information, see Tool parameters.

Design cloud architecture

This tool recommends architecture designs for cloud services, applications, and solutions — including file storage, banking, video streaming, e-commerce, SaaS, and more. It gathers requirements iteratively by asking 1–2 focused questions at a time, tracks a confidence score (0.0–1.0), and returns architecture guidance aligned with the Azure Well-Architected Framework. When the confidence score reaches 0.7 or higher, the tool stops asking follow-up questions and presents the architecture recommendation.

The tool covers all tiers: infrastructure, platform, application, data, security, and operations. Recommendations are conservative, actionable, and provide a high-level overview.

Example prompts include:

  • "Please help me design an architecture for a scalable file upload, storage, and retrieval service."
  • "Help me design an Azure-based ATM service architecture for user transactions and account management."
  • "I want to design a cloud app for ordering groceries with inventory and delivery tracking."
  • "How can I design an Azure cloud service to store, transcode, and serve videos to users?"
Parameter Required or optional Description
Answer Optional The user's response to the current question.
Confidence score Optional A value between 0.0 and 1.0 representing confidence in understanding requirements. When this reaches 0.7 or higher, nextQuestionNeeded should be set to false.
Next question needed Optional Whether another question is needed.
Question Optional The current question being asked.
Question number Optional Current question number.
State Optional The complete architecture state from the previous request as JSON. Tracks architecture components, tiers (infrastructure, platform, application, data, security, operations), requirements (explicit, implicit, assumed), and confidence factors.
Total questions Optional Estimated total questions needed.

Tool annotation hints:

Destructive: ❌ | Idempotent: ✅ | Open World: ❌ | Read Only: ✅ | Secret: ❌ | Local Required: ❌