Key Components of MCP

MCP defines three primary interaction types:

1. Tools - Functions the model can call (e.g., creating tickets, updating records)
2. Resources - Data or files the model can access (e.g., documents, database records)
3. Prompts - Reusable templates for specialized tasks

Architecture:

• MCP Client: The AI agent that makes requests (typically an LLM)
• MCP Server: Services that expose tools, data sources, or capabilities
• Transport Layer: Usually HTTP/JSON-RPC, with streaming capabilities

An AI using MCP can dynamically discover available capabilities, then decide which tools to use based on the task at hand. The protocol enables seamless data exchange and function calling in a standardized format.