MCP テストサーバー

What is MCP テストサーバー?

MCP テストサーバー is a Python server that implements the Model Context Protocol (MCP) to provide a Retrieval Augmented Generation (RAG) system using OpenAI's vector store.

How to use MCP テストサーバー?

Initialize the project with uv init, create a virtual environment, install dependencies with uv sync, and set the OPENAI_API_KEY environment variable in a .env file. The server exposes two tools: create_vector_db_from_directory and query_vector_db. It can be integrated with the Claude desktop app by adding a configuration entry pointing to the server's directory and the command uv run server.py.

Key features of MCP テストサーバー

• Creates a vector store from a local directory of files.
• Supports .txt, .pdf, .docx, and .md file formats.
• Queries the vector store to retrieve relevant information.
• Provides two MCP tools: creation and search.

Use cases of MCP テストサーバー

• Building a searchable knowledge base from project documentation.
• Enabling a chatbot to answer questions based on local documents.
• Creating a RAG pipeline with OpenAI’s vector store for fast retrieval.

FAQ from MCP テストサーバー

What does the server do?

It builds a vector store from files in a specified directory using OpenAI’s API and allows you to query that store for relevant content.

What are the runtime requirements?

Python 3.11 or higher and an OpenAI API key are required.

Does the server send my data to OpenAI?

Yes, the server uses the OpenAI vector store API, so file content is sent to OpenAI during vector store creation.

How can I test the server locally?

You can use the MCP Inspector by running npx @modelcontextprotocol/inspector uv run server.py and interact with the tools through its web interface.

How do I integrate this server with Claude?

Add a JSON entry in Claude desktop app’s config file with the server name, the uv command, and the absolute path to the project directory containing server.py.