Artificial Intelligence (AI) Agents and Model Context Protocol (MCP)
About Model Context Protocol (MCP)
Section titled “About Model Context Protocol (MCP)”Main article: Develop AI agents on Azure - Agents in Azure AI Solutions - see section of description of MCP and MCP processes
MCP and Agents
Section titled “MCP and Agents”Source: MCP vs. RAG: How AI Agents & LLMs Connect to Data - IBM - YouTube with Melissa Hadley, MCP vs API: Simplifying AI Agent Integration with External Data - IBM - YouTube with Martin Keen
MCP allows agents to act by accessing systems and data. Agent can:
- Discover available systems, data sources using MCP:
- Tools
- Application Programming Interface (API)
- Resources
- Prompt templates
- Understand user request, plan what tools to use
- Execute using tools from MCP server
- Integrate information and systems
Differences in API and MCP and Strengths of Both
Section titled “Differences in API and MCP and Strengths of Both”| Use Case | MCP | API |
|---|---|---|
| Purpose | Built for AI | General purpose |
| Discovery | Dynamic on connection | Fixed on releases |
| Interface | Same protocol, pattern | Unique |
MCP works well with APIs underneath.
To use the strengths of both, have AI use MCPs to discover API usage and other systems.
Tools and Processes in Context Engineering
Section titled “Tools and Processes in Context Engineering”If an agent must follow a process, use a skills. The skill can define using MCP for example to access and manage data.