# Stats Compass [Health: Active]

**Category:** 🧮 Data Science Tools  
**Repository:** https://github.com/oogunbiyi21/stats-compass-mcp  
**GitHub Stars:** 17  
**Views:** 5  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/stats-compass

## Description
50+ pandas-powered tools for data loading, cleaning, visualization, and ML workflows

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `uvx` (confidence: high):

```json
"mcpServers": {
  "stats-compass": {
    "command": "uvx",
    "args": ["stats-compass-mcp"],
    "env": {
      "STATS_COMPASS_PORT": "",
      "STATS_COMPASS_SERVER_URL": "",
      "STATS_COMPASS_MAX_UPLOAD_MB": ""
    }
  }
}
```

**Requires environment variables:** `STATS_COMPASS_PORT`, `STATS_COMPASS_SERVER_URL`, `STATS_COMPASS_MAX_UPLOAD_MB` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What Stats Compass MCP server does

Stats Compass MCP server gives an MCP-compatible AI client a collection of more than 50 tools for common data-science tasks. It belongs to the 🧮 Data Science Tools category and is intended for workflows that move from loading data through analysis and modeling.

The available areas include CSV and Excel loading, sample datasets, and DataFrame listing. Cleaning operations cover null handling, imputation, duplicate removal, and outlier handling. Transformation tools support filtering, grouping, pivoting, encoding, and adding columns. For analysis, the server provides descriptive statistics, correlation checks, hypothesis tests, and data-quality work. Visualization options include histograms, scatter plots, bar charts, ROC curves, and confusion matrices. Machine learning workflows include classification, regression, and time-series forecasting.

## How it works

In local mode, an AI client can ask Stats Compass to load a file by path or locate data files in a folder. The resulting data can then be passed through cleaning, transformation, exploratory analysis, visualization, or modeling operations. Prompts beginning with “Use stats compass to...” are recommended because some clients may otherwise write Python or use shell commands instead of selecting the server’s tools.

Stats Compass MCP server can also run as a remote HTTP service. The service exposes an MCP endpoint and provides a browser upload flow for files. Results such as exported DataFrames, plots, and trained models can be downloaded through generated links. This mode is suited to Docker deployments or setups where multiple clients connect to one running service.

## Setup and configuration

Install the Python package with `pip install stats-compass-mcp`. The project requires Python 3.11 or newer. Its installer can configure Claude Desktop with `stats-compass-mcp install --client claude` or VS Code with `stats-compass-mcp install --client vscode`. Claude Code can add the server with `claude mcp add stats-compass -- uvx stats-compass-mcp run`.

For remote use, start the service with `stats-compass-mcp serve --port 8000`. VS Code can connect to `http://localhost:8000/mcp`; Claude Desktop can use `mcp-proxy` to bridge to the streamable HTTP endpoint. Docker is also supported, including a published port and the `STATS_COMPASS_SERVER_URL` setting.

The configurable environment variables are `STATS_COMPASS_PORT`, which defaults to `8000`, `STATS_COMPASS_SERVER_URL`, which defaults to `http://localhost:8000`, and `STATS_COMPASS_MAX_UPLOAD_MB`, which defaults to `50`.

## Limitations and notes

Stats Compass MCP server’s documented client support includes Claude Desktop, VS Code Copilot, Claude Code CLI, and Cursor. GPT and Gemini support is marked partial. Remote operation requires the upload workflow for bringing files into the service, while local operation expects paths or folders accessible to the running setup.

The first Claude Code connection may fail while `uvx` downloads the package. Disabling and re-enabling the server in the MCP settings is documented as a recovery step. The server is released under the MIT license.

_Full upstream README: https://allmcps.com/mcp/stats-compass/readme_