VitalDB

You will set up a Model Context Protocol (MCP) server to perform advanced, natural-language–driven analytics on VitalDB data. This server lets you filter, batch analyze, correlate signals, compare groups, visualize results, detect anomalies, and export findings, all through an MCP client without writing complex data queries.

How to use

Connect your MCP client to the VitalDB MCP Server and start issuing natural-language requests that trigger the server’s advanced analytics. You can ask the server to filter cases by statistics, run batch analyses on multiple cases or tracks, compute correlations, compare two groups with automatic t-tests, create rich visualizations, detect anomalies, and export results to CSV. Your workflow might look like this: select one or more cases, choose the signals (tracks) you want to analyze, specify the statistical or time-window criteria, and then review the generated statistics, charts, and CSV exports. The system supports unlimited overlays or subplots for visual comparison and can render scatter plots with regression lines, heatmaps over time, and distribution charts.

How to install

Prerequisites you need before installation: a Python environment and access to the internet. You will install dependencies, configure the MCP server, and then run the server locally so your MCP client can connect.

cd /Users/yeze-macbookair-2025/workspaces/vitaldb-mcp-server
./install.sh

# Optional: verify Python environment if needed
python --version
python -m venv venv
source venv/bin/activate
pip install -r requirements.txt

# Configure MCP client (Claude Desktop example)

{ "mcpServers": { "vitaldb": { "command": "/Users/yeze-macbookair-2025/workspaces/vitaldb-mcp-server/venv/bin/python", "args": ["/Users/yeze-macbookair-2025/workspaces/vitaldb-mcp-server/server.py"] } } }

# Start the server (if you run it directly as a stdio MCP)
# This example uses the runtime command already configured for Claude Desktop

# Restart Claude Desktop after updating the config

Additional content

The VitalDB MCP Server supports many powerful tools and visualization options. You can discover and use the following capabilities to perform sophisticated analyses across many cases and tracks, with results you can export for further use.

Available tools

find_cases

Search for cases based on a track and metadata filters to assemble the dataset you will analyze.

get_case_info

Retrieve detailed information about a specific case for context before analysis.

search_available_tracks

List all tracks available for analysis so you can select signals to include.

filter_cases_by_statistics

Apply statistical criteria (e.g., mean, median, standard deviation, skewness, kurtosis) and simple expressions ( >100, >=80, <60, 50-100 ) to select cases.

batch_analyze_cases

Analyze multiple cases in one run, with support for multiple tracks per case.

analyze_correlation

Compute correlation coefficients between two signals and visualize overall and per-case relationships.

compare_groups

Perform statistical comparisons between two groups, including automatic t-tests and p-value interpretation.

detect_anomalies

Identify outliers using Z-score and IQR methods, with visualization of outlier positions.

time_window_analysis

Analyze data within defined time windows to study changing patterns.

export_to_csv

Save analysis results to CSV, supporting multiple cases and tracks.

plot_multiple_cases

Create plots that overlay or subplot many cases without a fixed limit.

plot_distribution

Visualize distributions with histograms, box plots, and violin plots.

plot_scatter_correlation

Plot scatter plots with optional regression lines to show relationships between signals.

plot_heatmap

Generate heatmaps showing signal value changes over time across tracks.