MCP client

The MCP client is the application-side component of the Model Context Protocol (MCP): it securely requests real-time enterprise data and executes external tools for Large Language Models (LLMs) and AI agents.

The MCP client is your secure bridge for grounding AI: it's the consumer-side application that initiates requests to an MCP server, ensuring LLMs operate on live, governed data. This client-server architecture, often leveraging Server-Sent Events (SSE) over HTTPS, facilitates three core functions. First, it enables real-time LLM grounding by retrieving specific context from data silos. Second, it orchestrates tool execution, allowing AI agents to securely trigger actions (e.g., updating a Salesforce CRM record). Third, it enforces strict data governance and privacy controls, ensuring only authorized information is returned to the AI application.

https://modelcontextprotocol.io/