MCP Server

This is a simple server application built using FastMCP, designed to handle various tools for retrieving information such as weather forecasts, web searches, Wikipedia queries, access to calendars, and more. Using these tools a local small LLM can be used with more recent data.

Provided Tools

• search_web: Performs internet searches to query unknown or recent data using DuckDuckGo.
• search_wikipedia: Queries Wikipedia for more detailed information on specific topics.
• read_pdf_file: Extract information from a local PDF file.
• get_available_calendars: Lists all available Google calendars. (*)
• get_calendar_entries: Lists all calendar entries for a specific calendar for the next N days. (*)
• get_current_weather: Retrieves the current weather forecast for a given location.
• get_current_date: Returns the current date in a formatted string.
• get_current_location: Returns the fixed location of the user (currently set to Mainz, Germany).
• ask_genai_knowledge_base: Passes the user's question to the RAG knowledge base.

(*) Please note, that you need to create a valid Google access token before using the tool for the first time, see https://developers.google.com/identity/protocols/oauth2?hl=en for details. If you run this server in a Docker container you need to create the token on the host and copy it onto the image.

Requirements

• Python 3.x; please note, that at the time of writing Python 3.14 is not compatible with ChromaDB thus Python 3.12 is used instead
• FastMCP
• ChromaDB, modify utils.config.py for configuration of host and port as ChromaDB is supposed to be available in a separate docker container by default
• OpenAI compatible LLM, modify utils.config.py if necessary
• PyMuPDF
• Other dependencies as specified in requirements.txt

Installation

1. Clone or download this repository.
2. Ensure all required dependencies are installed.

pip install fastmcp requests ddgs beautifulsoup4 chromadb wikipedia colorama langchain-text-splitters pymupdf google-api-python-client google-auth google-auth-oauthlib google-auth-httplib2

3. Run the server:

python mcp_server.py

Usage

The MCP Server runs on http://0.0.0.0:8002/mcp. You can interact with the server using the provided tools.

Docker

The server can be run in a docker container. A docker file as well as a setup script create_docker_image.sh is available.

Naming

The name of the MCP server "Sammelsurium" is German and translates roughly to "hodgepodge", which is intended to show that it is a collection of different tools.