RMCP

RMCP enables you to turn natural language conversations into comprehensive statistical analyses. By connecting an AI assistant or MCP client to a ready-to-use statistical engine, you can perform regression, time series forecasting, machine learning, hypothesis testing, data transformation, and visualization through plain language prompts. It supports a broad ecosystem of R packages and provides production-ready HTTP transport or local stdio operation for Claude Desktop, Claude Web, or other MCP clients.

How to use

Interact with RMCP through your MCP client by opening a session and asking for statistical analyses in everyday language. You can request regression and econometric models to quantify effects, time series forecasts and decompositions, clustering and other machine learning tasks, as well as standard statistical tests and descriptive analyses. RMCP returns formatted results, accompanying markdown tables, and inline visualizations to help you interpret outputs directly in your conversational interface.

How to install

Prerequisites: you need Python 3.10+ installed on your system.

Install the RMCP package from PyPI and set up HTTP transport if you plan to use the web-based MCP interface.

pip install rmcp
# Optional: enable HTTP transport support
pip install rmcp[http]

Additional sections

Claude Desktop integration allows you to run RMCP from the Claude Desktop MCP configuration by providing a simple stdio command that RMCP can execute locally.

{
  "mcpServers": {
    "rmcp": {
      "command": "rmcp",
      "args": ["start"]
    }
  }
}

HTTP server and local testing

If you want to use a remote MCP endpoint, you can access RMCP over HTTP at the production URL and test health and initialization calls from your environment.

Production server and docs endpoints are available at these URLs.

# Health check
curl https://rmcp-server-394229601724.us-central1.run.app/health

# Initialize MCP session (example; use your client to start a session)
curl -X POST https://rmcp-server-394229601724.us-central1.run.app/mcp \
  -H "Content-Type: application/json" \
  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"test-client","version":"1.0"}}}'

Security and reliability notes

RMCP uses a curated set of R packages drawn from CRAN task views, with a multi-tier security assessment to minimize exposure to risky dependencies. Review your environment permissions when loading datasets and running external analyses, and validate results in controlled pipelines.

Troubleshooting

If you encounter issues with the MCP connection or analysis results, verify that your RMCP transport is reachable and that the environment has the required R packages installed. Use the health endpoint to confirm availability and ensure your client is configured with the correct MCP server details.

Examples and workflows

You can request practical analyses such as ROI analysis from marketing spend, Okun’s Law tests in macro data, or churn prediction models from customer data. Provide clear inputs and target outcomes in natural language, and RMCP will select appropriate models, run analyses, and present interpretable results.

Available tools

regression

Perform linear and logistic regression, including panel data and instrumental variables as appropriate; produce coefficients, p-values, and model diagnostics.

time_series

Fit and forecast time series models such as ARIMA, decompose series, and assess stationarity; generate forecasts with confidence intervals.

machine_learning

Apply clustering, decision trees, random forests, and other ML methods to your data; provide interpretable model outputs.

statistical_tests

Run T-tests, ANOVA, chi-square tests, normality assessments, and other hypothesis tests with clear conclusions.

data_analysis

Compute descriptive statistics, correlations, outlier checks, and dataset summaries.

data_transformation

Standardize data, create lag/lead variables, and perform winsorization or other data prep steps.

visualizations

Create inline plots and visualizations suitable for Claude or other MCP clients to accompany results.

file_operations

Load data from CSV, Excel, JSON with validation checks before analysis.

nl_features

Support natural language-driven formula construction and example datasets for quick prototyping.