The Copilot Connection CCDev 2 - what you need to think about with MCP
Jul 29, 2025
In this discussion, Garry Trinder, a Developer Advocate at Microsoft, dives into the Model Context Protocol (MCP) and its game-changing potential for developers. He highlights the challenges of integrating MCP, from enhancing documentation to establishing governance. Trinder stresses the importance of crafting user-friendly materials and the adaptability of MCP servers for efficient software development. The conversation also delves into practical strategies for getting started with MCP while unlocking significant productivity gains in AI applications.
AI Snips
Chapters
Transcript
Episode notes
MCP Adds Context To LLMs
- Garry Trinder explains MCP supplies structured tools and context to LLMs so models can choose and invoke capabilities.
- Build an MCP server to expose named tools and descriptions that ground model reasoning.
MCP Supports Local Functions And APIs
- Garry Trinder stresses MCP is not limited to API calls and can run functions, query file systems, or access databases.
- Use MCP to expose non-API capabilities uniformly to any client that supports the protocol.
Agents Orchestrate Multi-Step Tool Use
- Garry Trinder describes agents iterating through multiple tools to build plans and then execute actions using MCP servers.
- That multi-step reasoning composes documentation search, code inspection, and file edits into a single workflow.