Model Context Protocol (MCP)#
The Model Context Protocol (MCP) is an open-source standard protocol that allows Large Language Models (LLMs) to interact with external tools and data sources through structured API calls.
MCP serves as a standardization layer for AI applications to communicate effectively with external services such as tools, databases and predefined templates. Think of MCP as a "USB-C port" for AI applications - it provides a standardized way for various tools, platforms, and data sources to connect to AI models.
Architecture#
MCP operates through a client-server architecture:
- MCP Hosts: Applications like Claude Desktop, IDEs, or AI tools that wish to access data via MCP
- MCP Clients: Protocol clients that maintain 1:1 connections with MCP servers
- MCP Servers: Lightweight services that expose capabilities (tools, resources, prompts) via the standardized protocol
Core Capabilities#
MCP supports three main types of capabilities:
- Tools: Functions that can be invoked with structured inputs
- Resources: Data sources that can be read (files, databases, etc.)
- Prompts: Reusable prompt templates with parameters
With LlamaIndex#
MCP introduces a compelling alternative to vector indexing for some use cases. Instead of crawling Salesforce data into a vector store for example, you can expose Salesforce as a live queryable tool.
This is particularly valuable for:
- Real-time Data: Get fresh data directly from source systems at query time
- Structured Queries: Handle complex queries requiring relationships and logic
- Action-Taking: Execute operations beyond just information retrieval