mcp-openai-gemini-llama-example

# How to use Anthropic MCP Server with open LLMs, OpenAI or Google Gemini This repository contains a basic example of how to build an AI agent using the Model Context Protocol (MCP) with an open LLM (Meta Llama 3), OpenAI or Google Gemini, and a SQLite database. It's designed to be a simple, educational demonstration, not a production-ready framework. OpenAI example: https://github.com/jalr4ever/Tiny-OAI-MCP-Agent ## Setup This code sets up a simple CLI agent that can interact with a SQLite database through an MCP server. It uses the official SQLite MCP server and demonstrates: * Connecting to an MCP server * Loading and using tools and resources from the MCP server * Converting tools into LLM-compatible function calls * Interacting with an LLM using the `openai` SDK or `google-genai` SDK. ## How to use it * Docker installed and running. * Hugging Face account and an access token (for using the Llama 3 model). * Google API key (for using the Gemini model). ### Installation 1. Clone the repository: ```bash git clone https://github.com/philschmid/mcp-openai-gemini-llama-example cd mcp-openai-gemini-llama-example ``` 2. Install the required packages: ```bash pip install -r requirements.txt ``` 3. Log in to Hugging Face ```bash huggingface-cli login --token YOUR_TOKEN ``` ## Examples ### Llama 3 Run the following command ```bash python sqlite_llama_mcp_agent.py ``` The agent will start in interactive mode. You can type in prompts to interact with the database. Type "quit", "exit" or "q" to stop the agent. Example conversation: ``` Enter your prompt (or 'quit' to exit): what tables are available? Response: The available tables are: albums, artists, customers, employees, genres, invoice_items, invoices, media_types, playlists, playlist_track, tracks Enter your prompt (or 'quit' to exit): how many artists are there Response: There are 275 artists in the database. ``` ### Gemini Run the following command ```bash GOOGLE_API_KEY=YOUR_API_KEY python sqlite_gemini_mcp_agent.py ``` ## Future plans I'm working on a toolkit to make implementing AI agents using MCP easier. Stay tuned for updates!