home / mcp / cocktails mcp server

Cocktails MCP Server

Delivers cocktail recommendations using a RAG pipeline with local FAISS indexing and Groq-powered LLMs.

Installation

Add the following to your MCP client configuration file.

Configuration

{
  "mcpServers": {
    "00200200-cocktails-rag-mcp": {
      "command": "uv",
      "args": [
        "run",
        "--with",
        "faiss-cpu",
        "--with",
        "fastmcp",
        "--with",
        "jq",
        "--with",
        "langchain",
        "--with",
        "langchain-community",
        "--with",
        "langchain-groq",
        "--with",
        "langchain-huggingface",
        "--with",
        "pandas",
        "--with",
        "python-dotenv",
        "--with",
        "sentence-transformers",
        "fastmcp",
        "run",
        "/ABSOLUTE/PATH/TO/src/mcp/server.py:mcp"
      ],
      "env": {
        "GROQ_API_KEY": "your_groq_api_key_here"
      }
    }
  }
}

You can run a dedicated MCP server that uses Retrieval-Augmented Generation to recommend cocktails. This server exposes an MCP endpoint you can connect to with your MCP client, handles data through a local vector store, and uses Groq’s API for language modeling. It’s designed to help you get quick, accurate cocktail suggestions built from a local dataset with efficient retrieval and ranking.

How to use

To use the Cocktails RAG MCP Server, connect your MCP client to the local server instance. Run the MCP in stdio mode to start a local server process that listens for requests from your client and returns cocktail recommendations powered by a RAG pipeline.

How to install

Prerequisites you need on your machine before installation:

Python 3.11 or newer
uv package manager (install from https://docs.astral.sh/uv/getting-started/installation/)

Step-by-step commands to set up and run the server locally:

# Optional: clone the project repository
# git clone https://github.com/00200200/cocktails-rag-mcp.git
# cd cocktails-rag-mcp

# Copy environment template and edit GROQ key
cp .env.example .env
# Edit .env to add GROQ_API_KEY
nano .env

# Install dependencies via UV
uv sync

# Pre-download models (embeddings and reranker)
uv run python -c "from src.rag.rag import RAG; RAG(); print('Models downloaed!')"

# Start the MCP server locally in stdio mode
uv run python src/mcp/server.py

Additional setup and configuration

If you want to run the server in a pre-configured environment for Claude Desktop or a specific setup, you can adopt the provided MCP JSON configuration to register the server with your client. The following configuration demonstrates a stdio MCP setup using UV and FastMCP, including the required environment variable for GROQ.

{
  "mcpServers": {
    "cocktails": {
      "command": "uv",
      "args": [
        "run",
        "--with","faiss-cpu",
        "--with","fastmcp",
        "--with","jq",
        "--with","langchain",
        "--with","langchain-community",
        "--with","langchain-groq",
        "--with","langchain-huggingface",
        "--with","pandas",
        "--with","python-dotenv",
        "--with","sentence-transformers",
        "fastmcp",
        "run",
        "/ABSOLUTE/PATH/TO/src/mcp/server.py:mcp"
      ],
      "env": {
        "GROQ_API_KEY": "your_groq_api_key_here"
      }
    }
  }
}

Available tools

RAG pipeline

Orchestrates retrieval from a FAISS vector store, reranking of candidate results, and generation of natural language cocktail recommendations.

Embeddings loading

Loads local embeddings for cocktail data to enable semantic search within the RAG workflow.

FAISS indexing

Manages the local FAISS index for fast vector similarity search against cocktail data.