Model Context Protocol (MCP)

What is Model Context Protocol (MCP)?

Model Context Protocol (MCP) is a client-server protocol that enables communication between large language models (LLMs) and external tools or data sources. It uses JSON‑RPC 2.0 for messaging and was introduced by Anthropic.

How to use Model Context Protocol (MCP)?

Implement an MCP server using FastMCP or the Low‑Level APIs from the MCP Python SDK. Servers can be run over SSE (HTTP) or stdin. The README provides full Python examples for both server and client (SSE and stdio). For instance, a FastMCP server can be served with mcp.run(transport="sse") or wrapped with Starlette.

Key features of Model Context Protocol (MCP)

• Client‑server architecture using JSON‑RPC 2.0
• Three server capabilities: Tools, Resources, and Prompts
• Supports SSE over HTTP and stdin transport layers
• FastMCP high‑level API for quick server development
• Standardized initialization and discovery flow
• Built‑in error codes for JSON‑RPC

Use cases of Model Context Protocol (MCP)

• Providing an LLM with an assessment quiz (Resource)
• Letting an LLM call a triage function to determine user level (Tool)
• Generating dynamic prompt templates for an LLM (Prompt)
• Any scenario where an LLM needs structured access to external data or logic

FAQ from Model Context Protocol (MCP)

What is Model Context Protocol (MCP)?

MCP is a protocol that establishes a communication channel between large language models and external tools or data sources. It follows a client‑server architecture and uses JSON‑RPC 2.0 for messaging.

What capabilities can an MCP server provide?

An MCP server can provide three capabilities: Tool (a function that performs logic or side effects), Resource (any type of data, e.g. images, files), and Prompt (a prompt template for the client).

How does the transport layer work?

MCP supports two transport methods: SSE over HTTP (for network communication) and stdin (for local communication). The transport type is set when starting the server.

What is FastMCP?

FastMCP is a high‑level Python API from the MCP Python SDK that simplifies server implementation. It provides decorators to define @mcp.resource, @mcp.tool, and @mcp.prompt easily.

How do I implement an MCP client?

Clients can be implemented using the MCP Python SDK with either sse_client (for SSE servers) or stdio_client (for stdio servers). Both examples are provided in the README, including session initialization, tool listing, and tool calling.