Add additional MCP tools for 'Azure Well-Architected Framework'

[KarishmaGhiya]

Service / Tool Name

Tool name : Azure.Mcp.Tools.WellArchitectedFramework

Contacts

Arun Rabindar (arunrab), Sriram Kakara (skakara), Jon Brockett (jbrockett), Michael Ren (mren)

Intended Agent Scenarios

Issue

Add a new MCP tool for Azure Well-Architected Framework

Scenarios

This tool provides quick and efficient access to any skill that needs to refer to the 'Azure Well-Architected Framework' for architecture guidelines.
Ex: We have an Azure infrastructure planning skill which needs to refer to WAF to come up with the best possible architecture.

Commands

Some sample commands we'll expose :

• WorkloadGuideGetCommand(workloadClass)
• ServiceGuideGetCommand(serviceName)
• RecommendationsListCommand(pillar)
• RecommendationGetCommand(recommendationId)
• DesignPrinciplesGetCommand(pillar)
• AnalyzeInfraCommand(infraConfig, [pillars])

Why have it as a MCP tool and not instruct the agent in the skill to directly refer to WAF website?

WAF documentation is extensive across several pillars (cost, security, reliability, performance, operations) and several kinds of Azure services. Asking the LLM to refer to WAF at runtime will :

• Increase and dilute its context window. Instead, this MCP tool enables the agent to focus on the guidelines for a specific class of workload, with specific pillars and specific services. Focus vs bloat.
• Saves tokens
• Better runtime performance, as WAF documentation will already be LLM analyzed and summarized once, and returned as plain text by the MCP tool. Client LLM need not analyze for every invocation during run-time.
• Also, only ~60% or so WAF guidelines apply to agents that are trying to architect. Around ~40% of WAF guidelines (like CO:01, SE:02, RE:01, OE:01, PE:11 and several others) apply only to a human cloud solution architect and not to an agent. So we will filter those out to make the agent more focused.

Timeline

Will be for v3 GA (starting April)

[github-actions]

Thanks for your feedback! @KarishmaGhiya is looking into it.