ML Research MCP Server

ML Research MCP

A comprehensive Model Context Protocol (MCP) server providing research productivity tools for machine learning researchers and developers.

Overview

ML Research MCP is an extensible platform that provides AI assistants with powerful tools for scientific research workflows. Built on the Model Context Protocol, it enables seamless integration with AI applications like Claude Desktop to automate and enhance various research tasks.

Current Status: Phase 1 - Data Visualization Roadmap: Image generation, presentation tools, literature management, and more

Vision & Roadmap

This project aims to be a comprehensive research assistant covering the entire ML research lifecycle:

✅ Phase 1: Data Visualization (Current)

• Scientific plotting with publication-quality output
• Statistical analysis visualizations
• 2D data representations (heatmaps, contours)
• Multiple export formats (PNG, PDF, SVG)

🚧 Phase 2: Image & Figure Generation (Planned)

• AI-powered figure generation using diffusion models
• Diagram creation for architecture illustrations
• Chart enhancement with intelligent styling
• Multi-panel figure composition

🚧 Phase 3: Presentation Tools (Planned)

• Slide generation from research content
• Poster creation for conferences
• Automated layout optimization
• Template management for institutional branding

🚧 Phase 4: Research Management (Future)

• Literature search and summarization
• Citation management and formatting
• Experiment tracking and versioning
• Collaboration tools for team projects

Current Features (Phase 1)

Data Visualization Tools

Basic Plots

• plot_line - Time series and continuous data visualization
• plot_scatter - Multi-dimensional scatter plots with size/color mapping
• plot_bar - Categorical comparisons (vertical/horizontal)

Statistical Visualizations

• plot_histogram - Distribution analysis with density estimation
• plot_box - Statistical summaries and outlier detection
• plot_violin - Detailed distribution shapes with KDE

2D Representations

• plot_heatmap - Matrix visualization with annotations
• plot_contour - 3D data in 2D with contour lines
• plot_pcolormesh - Fast pseudocolor plots for large datasets

Technical Highlights

• Publication-quality output via UltraPlot
• High-performance data handling with Polars
• Flexible input from CSV, JSON files or direct data
• Vector & raster formats (PDF, SVG, PNG)
• Type-safe with comprehensive validation
• Well-tested with 48 passing tests

Requirements

• Python 3.12+
• uv package manager
• MCP-compatible client (Claude Desktop, VSCode, etc.)

Installation

Quick Start with Claude Code

Add the server to Claude Code with a single command:

claude mcp add-json "ml-research" \
  '{"command":"uvx","args":["--from","git+https://github.com/nishide-dev/ml-research-mcp","ml-research-mcp"]}'

Verify installation:

claude mcp list

Manual Installation for MCP Clients

Add to your MCP client configuration (e.g., ~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "ml-research": {
      "command": "uvx",
      "args": [
        "--from",
        "git+https://github.com/nishide-dev/ml-research-mcp",
        "ml-research-mcp"
      ]
    }
  }
}

Direct Execution with uvx

Run the server directly without installation:

# From GitHub (recommended)
uvx --from "git+https://github.com/nishide-dev/ml-research-mcp" ml-research-mcp

# From local directory (for development)
cd /path/to/ml-research-mcp
uvx --from . ml-research-mcp

For Developers

Clone and set up development environment:

git clone https://github.com/nishide-dev/ml-research-mcp.git
cd ml-research-mcp
uv sync

# Run in development mode
uv run ml-research-mcp

Quick Start

Using with Claude Desktop

After installation, you can ask Claude:

"Create a line plot showing temperature over time from experiment.csv"

"Generate a heatmap of the correlation matrix and save as PDF"

"Plot a scatter chart with x=[1,2,3,4], y=[2,4,6,8], sized by [10,20,30,40]"

Using as a Library

from ml_research_mcp.tools.plot_basic import plot_line

# Generate publication-quality plot
image = plot_line(
    x=[1, 2, 3, 4, 5],
    y=[1, 4, 9, 16, 25],
    style={"title": "Quadratic Function", "xlabel": "X", "ylabel": "Y²"},
    output={"format": "pdf", "width": 20, "height": 15, "dpi": 300}
)

# Save to file
with open("plot.pdf", "wb") as f:
    f.write(image)

Example Gallery

Real examples generated with ML Research MCP: