MCP Analytics is a statistical analysis platform integrated into AI chat clients via the MCP (Model Context Protocol) protocol. Users only need to provide data and a question, and the platform dispatches a set of specialized agents (spec drafting, building, validation, repair, deployment) to generate a customized analysis for that data and return a citable, interactive report.
Purpose
MCP Analytics is designed to give AI assistants professional statistical analysis capabilities. Users can directly perform statistical analysis, forecasting, and machine learning modeling on CSV files or live data sources in Claude, Cursor, Windsurf, or any MCP-compatible client, without writing code.
Main Capabilities
- Natural language analysis: Describe analysis needs in everyday language, and the platform automatically generates a customized analysis.
- Multiple depth levels: Provides four analysis depth levels (snapshot, json, brief, deck), from instant snapshots to full studies; the higher the depth, the more rigorous the statistical methods used (hypothesis testing, regression, diagnostics, etc.).
- Preset modules: Built-in modules such as t-tests, regression, churn analysis, customer segmentation, forecasting, customer LTV, A/B testing, time series, and survival analysis allow you to view a complete report in less than a minute.
- Data connections: Supports CSV upload, public URLs, and live OAuth connectors for Google Analytics 4 and Google Search Console (more connectors coming soon). After connecting, each rerun automatically pulls the latest data.
- Citable reports: One-click generation of APA / MLA / Chicago / BibTeX citations, suitable for papers, presentations, and regulatory documents.
- Reproducible: Fixed random seeds, Docker isolation, and named methods ensure that the same input always produces the same output.
- Source code embedding: Each report embeds the R source code so readers can run and verify it themselves.
- Personal library: All commissioned analyses are saved in a private library, can be queried from any MCP client, rerun on new data with one click, and shared as needed.
- Cross-library questions: The ask_library tool can pose a question to all delivered analyses and obtain a comprehensive answer with citations.
How to Use
- Get an API Key: Register for free at account.mcpanalytics.ai, get an API Key starting with mcp_, and receive 9,000 welcome credits, no credit card required.
- Install and configure: Supports stdio MCP clients such as Claude Desktop, Cursor, Windsurf, and Claude Code (requires Node.js 18+), and also supports Streamable HTTP transport. Configuration methods include editing claude_desktop_config.json, .cursor/mcp.json, or running an npx command.
- Start analyzing: After restarting the MCP client, ask questions in natural language, for example, "Upload sales.csv and identify revenue drivers" or "Forecast next quarter's sales."
Workflow: Upload data (datasets_upload) → Commission analysis (create_analysis, specifying the question, dataset, and depth level) → Track the build (build_status) → Get the report (reports_view / report_cards) → Rerun (run_analysis).
Use Cases
- Business data analysis: sales, customer, marketing, and forecasting analysis.
- Academic research: statistical analysis that needs to be citable and reproducible.
- Experimental design: A/B testing, hypothesis testing.
- Customer analysis: RFM segmentation, churn prediction, LTV modeling.
- Time series forecasting: trend detection, seasonality analysis.
Limitations
- The core server is proprietary commercial software, and the API server code is maintained separately.
- Custom analysis creation is a paid event, and failed builds are not billed.
- Free tools can run in the browser, but full functionality requires an API Key.
- For data privacy, processing is temporary and not persisted, and analyses run in isolated Docker containers.
- Specific details such as pricing, credit consumption, and dataset size limits are not yet provided.
