MCP Test Server

🚀 MCP Test Server

This is a Python server using the Model Context Protocol (MCP) to provide a Retrieval Augmented Generation (RAG) system based on OpenAI vector storage.

✨ Features

This MCP server offers the following features:

1. Vector Database Creation: Build an OpenAI vector store from files (including text, PDF, DOCX, and Markdown files) in a specified directory.
2. Vector Database Search: Execute queries on the built vector database to obtain highly relevant information.

📋 System Requirements

• Python 3.11 or higher
• OpenAI API key

📦 Installation

1. Initialize the project and create a virtual environment

# Initialize the project
uv init mcp-test-server
cd mcp-test-server

# Create and activate the virtual environment
uv venv .venv
source .venv/bin/activate

2. Install the necessary packages

# Install the required packages
uv sync

3. Set environment variables

Create a .env file and add your OpenAI API key:

OPENAI_API_KEY="your-api-key"

💻 MCP Server Testing (Optional)

Test using MCP Inspector

Start the MCP Inspector to verify the server functionality:

# Start the MCP Inspector
npx @modelcontextprotocol/inspector uv run server.py

The steps to operate in the MCP Inspector are as follows:

1. Select "Connect" in the left pane.
2. Choose "Tools" in the middle tab.
3. Execute "List Tools".
4. Select the tool to verify.
5. Set the parameters in the right pane and then execute "Run Tool".
6. The response content will be displayed in the Result area (due to unicode escape, appropriate decoding is required for viewing).
7. After completion, terminate the Inspector's operation in the CLI to release the port.

🔗 Integration with Claude Application

Add the following settings to the configuration file of the Claude desktop application:

"test_calc": {
  "command": "uv",
  "args": [
      "--directory",
      "/abs/path/to/your/mcp-test-server",
      "run",
      "server.py"
  ]
}

Replace /abs/path/to/your/mcp-test-server with the actual project path.

After adding the configuration, restart the Claude application to use the MCP server features.

🛠️ Provided Tools

This MCP server provides the following tools:

1. create_vector_db_from_directory

Create a vector database from files in a specified directory.

Parameters:

• directory_path: The directory path of the files to be processed (required).
• vector_store_name: The name of the vector store (default: "local_knowledge").
• file_patterns: The patterns of the files to be processed (default: ["*.txt", "*.pdf", "*.docx", "*.md"]).

Return Value:

• Information about the processing result (status, message, statistics, etc.).

2. query_vector_db

Search the built vector database to obtain relevant results.

Parameters:

• query: The search query (required).
• vector_store_name: The name of the vector store to query (default: "local_knowledge").

Return Value:

• A list of relevant documents containing the query results.