Working notes · Azure & AWS
Grounding, agents and evaluation - with the numbers attached.
Notes from production deployments on Copilot Studio, Azure AI Search, Azure AI Foundry and Amazon Bedrock. Retrieval quality, agent routing, eval gates, and the invoices that follow - measured before claimed.
- 1
- Posts
- 4
- Platforms
- 2026
- Since
Threat Model - Model Context Protocol (MCP) in Microsoft Copilot Studio
A risk register for Copilot Studio agents that consume MCP servers: twenty-four risks across supply chain, tool integrity, identity, data protection, governance, operations and compliance, each with its mitigating controls, the platform feature that delivers them, and the residual rating that survives.
Read the piece →29 min · 5,871 words