What is Transformerlab Mcp

transformerlab-mcp is an MCP (Model Context Protocol) server that provides an interface for AI assistants to interact with the functionalities of Transformerlab, including model management, training, dataset operations, evaluation, and RAG capabilities.

Use cases

Use cases for transformerlab-mcp include developing AI assistants that require interaction with Transformerlab’s features, managing machine learning models, conducting training and evaluations, and utilizing RAG techniques for enhanced AI performance.

How to use

To use transformerlab-mcp, ensure that the Transformerlab API server is running, then start the MCP server using the provided script. Follow the installation instructions to set up the environment and dependencies.

Key features

Key features of transformerlab-mcp include model management tools, dataset management tools, training and fine-tuning capabilities, evaluation tools, and RAG functionalities, all wrapped in a user-friendly interface for AI assistants.

Where to use

transformerlab-mcp can be used in various fields such as AI development, machine learning research, data science, and any application requiring advanced model management and training capabilities.

Transformerlab MCP Server

An MCP (Model Context Protocol) server that provides an interface for AI assistants to interact with Transformerlab functionality.

Overview

This project implements an MCP server that wraps the Transformerlab API, allowing AI assistants like Claude to interact with Transformerlab’s features. This includes model management, training, dataset operations, evaluation, and RAG capabilities.

Prerequisites

• Python 3.10 or higher
• A running Transformerlab API server (default: http://localhost:8338)
• UV (Python package manager)
• MCP SDK
• Transformerlab client library

Project Structure

transformerlab-mcp/
├── pyproject.toml          # Project configuration
├── README.md               # Documentation
├── src/
│   └── transformerlab_mcp/
│       ├── __init__.py     # Package initialization
│       ├── client.py       # Client wrapper for Transformerlab
│       ├── server.py       # MCP server implementation
│       └── tools/          # Folder for tool implementation
│           ├── __init__.py
│           ├── models.py   # Model management tools
│           ├── datasets.py # Dataset management tools
│           ├── training.py # Training and fine-tuning tools 
│           ├── evaluation.py # Evaluation tools
│           └── rag.py      # RAG tools
├── install_dev.sh          # Script to install in development mode
└── run.sh                  # Script to run the server

Installation

1. Ensure UV is installed (if not, install it following the instructions at https://github.com/astral-sh/uv)

2. Create a Python environment and install dependencies:

   uv venv
   source .venv/bin/activate
   uv pip install "mcp[cli]" transformerlab-client
   

3. Install the package in development mode:

   ./install_dev.sh
   

Usage

1. Ensure Transformerlab is running and accessible at the configured URL (default: http://localhost:8338)

2. Start the MCP server:

   ./run.sh
   

3. To test the server with the MCP Inspector:

   mcp dev -m transformerlab_mcp.server
   

4. To use with Claude Desktop or other MCP clients, add the following to the Claude Desktop config file (typically located at ~/.config/Claude Desktop/claude_desktop_config.json on Linux or similar locations on other platforms):

   {
     "mcpServers": {
       "TransformerLabMCP": {
         "command": "uv",
         "args": [
           "--directory",
           "/absolute/path/to/transformerlab-mcp",
           "run",
           "-m",
           "transformerlab_mcp.server"
         ]
       }
     }
   }
   

Supported Features

The MCP server provides tools for:

• Model Management: Listing, downloading, and getting information about models
• Training & Fine-tuning: Starting and monitoring training jobs
• Dataset Management: Listing, importing, and getting information about datasets
• Evaluation: Running evaluations and retrieving results
• RAG: Adding documents, listing collections, and performing RAG queries

Configuration

The server uses the following environment variables:

• TRANSFORMERLAB_API_URL: The URL of the Transformerlab API server (default: http://localhost:8338)

License

AGPL-3.0 (matching Transformerlab’s license)