mcp - scaffold is a development sandbox for verifying the Model Context Protocol (MCP) server, providing support for local LLMs (such as LLaMA 7B) and cloud inference, including a chat interface and reference architecture.

This project implements a document retrieval server based on the Model Context Protocol (MCP), which can dynamically obtain the latest official documentation content of Python libraries for AI assistants. It supports libraries such as LangChain, LlamaIndex, and OpenAI, conducts efficient searches through the SERPER API, and uses BeautifulSoup to parse HTML content. The project is designed to be extensible, facilitating the addition of support for more libraries.

This project integrates the Yahoo Finance API with the LLaMA 3.2 3B model. Through the Model Context Protocol (MCP), the model can access real - time financial data, stock information, and market news while retaining all the original functions of LLaMA.

The project introduces how to quickly set up an MCP service based on Ollama and the Llama3.2 model, including environment preparation and running steps.

A lightweight MCP server that retrieves relevant documentation content from popular AI libraries such as LangChain, LlamaIndex, and OpenAI by combining web search and content parsing, providing language models with the function of dynamically querying the latest documentation.