Stats Compass — Claude Code MCP server: features and setup
Stats Compass gives an AI agent tools to load and analyze tables, make charts, and run ML tasks. Connect it when you want to work with data files through prompts.
claude mcp add --transport http stats-compass https://mcp.statscompass.io/mcpWhat Stats Compass does
Stats Compass exposes tools for loading CSV and Excel files, handling missing values and duplicates, and filtering or grouping tables. Ask for data summaries, correlations, or hypothesis tests; it can also create visualizations, train classification and regression models, and forecast time series.
Local use requires Python 3.11+ and the stats-compass-mcp package; the MCP client can launch it with uvx. In Claude Code, add the local server with claude mcp add stats-compass -- uvx stats-compass-mcp. The first connection can fail while uvx downloads the package, so toggle the server off and back on in MCP settings if that happens. Remote access binds to 127.0.0.1 by default; connecting from other machines requires configuring a token.
What it helps with
- load a CSV or Excel file and inspect its structure
- clean missing values, duplicates, and outliers
- filter or group a table and add a column
- test correlations or differences between groups
- train a classification model and visualize results
Example prompts for the agent
- Load
sales.csvand summarize the main patterns in its data. - Clean the table by filling or removing missing values, then save it as a CSV.
- Test whether scores differ between groups A and B, and make a chart of the results.
How to install Stats Compass
1Platform CLI: remote server
claude mcp add --transport http stats-compass https://mcp.statscompass.io/mcp2Platform CLI
claude mcp add stats-compass -- uvx stats-compass-mcpuvx runs the command named after the package; if the README names another one, use uvx --from stats-compass-mcp <command>
1Platform CLI: remote server
codex mcp add stats-compass --url https://mcp.statscompass.io/mcp2Platform CLI
codex mcp add stats-compass -- uvx stats-compass-mcpuvx runs the command named after the package; if the README names another one, use uvx --from stats-compass-mcp <command>
Config file
{
"mcpServers": {
"stats-compass": {
"url": "https://mcp.statscompass.io/mcp"
}
}
}{
"mcpServers": {
"stats-compass": {
"command": "uvx",
"args": [
"stats-compass-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from stats-compass-mcp <command>
Config file
{
"servers": {
"stats-compass": {
"type": "http",
"url": "https://mcp.statscompass.io/mcp"
}
}
}{
"servers": {
"stats-compass": {
"type": "stdio",
"command": "uvx",
"args": [
"stats-compass-mcp"
]
}
}
}uvx runs the command named after the package; if the README names another one, use uvx --from stats-compass-mcp <command>
1Platform CLI: remote server
gemini mcp add -t http stats-compass https://mcp.statscompass.io/mcp2Platform CLI
gemini mcp add stats-compass uvx stats-compass-mcpuvx runs the command named after the package; if the README names another one, use uvx --from stats-compass-mcp <command>
Step by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about Stats Compass
How do I install Stats Compass in Claude Code?
For a local connection, run claude mcp add stats-compass -- uvx stats-compass-mcp. Python 3.11+ is required.
What does Stats Compass do?
It gives an agent tools for loading and cleaning tables, exploring data, creating visualizations, and running machine-learning tasks.
Which file types can Stats Compass load?
Local workflows support CSV and Excel files. With a remote server, users can upload files through a browser.
