Minima MCP Server

Overview

Minima is a powerful on-premises RAG system that lets you search and query your local documents using AI. It offers three distinct operation modes to suit different needs:

1. Isolated Installation: Run completely on-premises with containers, free from external dependencies. All neural networks (LLM, reranker, embedding) operate on your cloud or PC, ensuring your data remains secure.

2. ChatGPT Integration: Query your local documents using ChatGPT app or web with custom GPTs. The indexer runs on your cloud or local PC, while ChatGPT serves as the primary LLM.

3. Anthropic Claude Integration: Use Anthropic Claude app to query your local documents via MCP. The indexer operates on your local PC, while Claude serves as the primary LLM.

Installation

Prerequisites

• Docker and Docker Compose
• Python 3.10 or higher (for MCP usage)
• uv package installed (for MCP usage)

Setting Up Environment Variables

1. Create a .env file in the project's root directory based on the provided .env.sample.
2. Configure the following variables:

LOCAL_FILES_PATH=/path/to/your/documents/
EMBEDDING_MODEL_ID=sentence-transformers/all-mpnet-base-v2
EMBEDDING_SIZE=768

For fully local installation, also add:

OLLAMA_MODEL=qwen2:0.5b
RERANKER_MODEL=BAAI/bge-reranker-base

For ChatGPT integration, add:

USER_ID=your-email@example.com
PASSWORD=your-password

Installation Methods

Docker Compose (Manual Installation)

Choose one of the following commands based on your preferred mode:

• Fully Local Installation:

  docker compose -f docker-compose-ollama.yml --env-file .env up --build
  

• ChatGPT Integration:

  docker compose -f docker-compose-chatgpt.yml --env-file .env up --build
  

• MCP Integration (for Anthropic Claude):

  docker compose -f docker-compose-mcp.yml --env-file .env up --build
  

Smithery Installation (for MCP usage)

For automatic installation with Claude Desktop:

npx -y @smithery/cli install minima --client claude

Accessing Minima

• Fully Local Installation: Navigate to cd electron, run npm install and npm start to launch the Minima electron app.
• Web Interface: Access the chat UI at http://localhost:3000
• ChatGPT Integration: Copy the OTP from the terminal where you launched Docker and use Minima GPT
• Claude Integration: Configure Claude Desktop by adding Minima to your Claude configuration file.

Configuration Details

Environment Variables Explained

• LOCAL_FILES_PATH: Root folder for indexing (recursive). Supports .pdf, .xls, .docx, .txt, .md, .csv files.
• EMBEDDING_MODEL_ID: Embedding model to use (Sentence Transformer models).
• EMBEDDING_SIZE: Embedding dimension provided by the model.
• OLLAMA_MODEL: Ollama LLM model ID (for local installation).
• RERANKER_MODEL: Reranker model (BAAI rerankers recommended).
• USER_ID: Email for ChatGPT authentication.
• PASSWORD: Password for ChatGPT authentication.

Claude Desktop Configuration

To use Minima with Claude Desktop, add the following to your Claude configuration file at /Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "minima": {
      "command": "uv",
      "args": [
        "--directory",
        "/path_to_cloned_minima_project/mcp-server",
        "run",
        "minima"
      ]
    }
  }
}

Replace /path_to_cloned_minima_project with the actual path to your cloned Minima repository.

Supported File Types

Minima can index and search through the following file types:

• PDF (.pdf)
• Excel (.xls)
• Word (.docx)
• Text (.txt)
• Markdown (.md)
• CSV (.csv)

All documents within the specified LOCAL_FILES_PATH folder and its subfolders will be indexed automatically.