# MCP Analytics — Full Reference > MCP Analytics is a statistical analysis and reporting platform. Upload CSV data or connect live sources, get interactive HTML reports with validated R analyses, AI insights, and exportable PDFs. Available via the account app (account.mcpanalytics.ai) and via the Model Context Protocol (MCP) for AI assistants. ## What Makes It Different MCP Analytics runs validated R scripts — the same statistical language used in academia, pharma, and finance. Unlike AI tools that generate Python code on the fly (different every run), our modules produce the same result every time. The AI provides insights and interpretation, not the methodology. Every report includes: - 6-16 interactive cards with Plotly charts (hover, zoom, sort) - AI-generated insights interpreting each visualization - The R code that produced the results (copy-to-clipboard) - A methodology description explaining the test, assumptions, and limitations - Citation format (APA/BibTeX) for academic use - One-click PDF export for presentations and board decks - A reproducibility statement guaranteeing identical results on repeat ## Pricing (Current) Purchasing is credit-based. No seat licenses — pay for what you run. **Free entry:** 2,000 welcome credits on signup, no card required. A Snapshot analysis costs 100 credits, so the free tier covers up to 20 instant analyses. **Pay As You Go:** buy credit packs, use anytime, credits never expire. Haiku AI insights. - $5 / 500 credits - $13 / 1,500 credits (13% off) - $40 / 5,000 credits (20% off) - $110 / 15,000 credits (27% off) - $325 / 50,000 credits (35% off) **Subscriptions** for volume users (42–45% off PAYG base rate, permanent report archive): - Pro: $99/month — 17,000 credits (≈ 5 module creations + 60 report runs), Sonnet AI - Business: $499/month — 90,000 credits, Sonnet + Opus AI, live connectors (GA4, GSC), unlimited users - Enterprise: custom contract, unlimited credits — contact sales **What things cost:** Snapshot 100 credits (~2 minutes) · report rerun ~200 credits · Atlas deep analysis 5,000 credits (30–45 minutes) · custom module creation priced by depth, charged only if the build succeeds. **Satisfaction guarantee:** Every report has feedback buttons on each card. Flag an issue — we fix and resend. Still not satisfied? Credits back, no questions asked. ## Analysis Capabilities ### Statistical Testing - t-test (independent, paired, one-sample) - ANOVA (one-way, with post-hoc tests) - Chi-square test of independence - Mann-Whitney U test (non-parametric) - Kruskal-Wallis test (non-parametric ANOVA) - Correlation analysis (Pearson, Spearman) ### Regression - Linear regression (OLS) - Ridge regression (L2 regularization) - Lasso regression (L1 regularization) - Elastic net regression - Logistic regression / classification ### Machine Learning - Random forest (classification and regression) - XGBoost (gradient boosting) - K-means clustering - DBSCAN clustering (density-based) - PCA (dimensionality reduction) - Isolation forest (anomaly detection) - Naive Bayes classification ### Time Series - ARIMA (autoregressive integrated moving average) - Simple trend analysis - Revenue time series forecasting ### Research Methods - ICC (Intraclass Correlation Coefficient) for inter-rater reliability - ANCOVA (treatment effects with covariates) - Cox proportional hazards (survival analysis) - Holm-Bonferroni correction (multiple comparisons) - Difference-in-differences (causal inference) ### Business Analytics - ROAS / Ad spend efficiency analysis - Marketing campaign attribution - Price elasticity analysis - MRR (Monthly Recurring Revenue) analysis - Churn prediction (subscription businesses) - RFM segmentation (customer value) - Product profitability analysis - Revenue forecasting - Cohort retention analysis ### Engagement & SEO - GA4 page engagement analysis - GA4 traffic source analysis - GSC ranking changes - GSC quick wins (high-impression, low-CTR pages) - Content performance analysis - Cross-platform content comparison - GSC portfolio analysis - GSC title A/B testing ### General Purpose - CSV auto-profiler (instant data overview) - Pivot summary tables - Dataset join operations - Generic categorical/numeric summary ## Free Tools Upload a CSV and get a real report — no signup required. Same analysis engine as paid tier. - [CSV Auto-Analyzer](https://mcpanalytics.ai/free/csv-analyzer): Instant data profiling - [Correlation Analysis](https://mcpanalytics.ai/free/correlation-analysis): Find relationships - [t-Test](https://mcpanalytics.ai/free/t-test): Compare two groups - [ANOVA](https://mcpanalytics.ai/free/anova): Compare multiple groups - [Linear Regression](https://mcpanalytics.ai/free/linear-regression): Predict outcomes - [Chi-Square Test](https://mcpanalytics.ai/free/chi-square): Test categorical associations - [Time Series Trend](https://mcpanalytics.ai/free/time-series-trend): Detect patterns Full catalog: https://mcpanalytics.ai/free/ ## How It Works 1. **Upload data** — drag-and-drop CSV, or connect GA4/GSC 2. **Pick analysis** — or let the AI agent recommend. 5-signal semantic matching (structural similarity, LLM description, LLM overview, column type coverage, category fit) finds the right module 3. **Get report** — interactive HTML with charts, tables, AI insights, R code, methodology, PDF export 4. **Build knowledge** — every result gets embedded in a shared semantic space. Search by meaning across all past analyses ## MCP Integration MCP Analytics is an MCP server. Install in any compatible AI client: ```json { "mcpServers": { "mcpanalytics": { "command": "npx", "args": ["-y", "@mcp-analytics/mcp-analytics"], "env": { "MCP_ANALYTICS_API_KEY": "YOUR_API_KEY" } } } } ``` Supported clients: Claude Desktop, Cursor, Windsurf, Claude Code, any MCP-compatible agent. Four connection methods: 1. **npx stdio** (recommended) — `npx -y @mcp-analytics/mcp-analytics` with `MCP_ANALYTICS_API_KEY` env var 2. **Direct HTTP** — `POST https://api.mcpanalytics.ai/mcp/api-key` with X-API-Key header 3. **OAuth2** — `https://api.mcpanalytics.ai/auth0` for webapp integrations 4. **Public discovery** (no auth) — `https://api.mcpanalytics.ai/mcp/discover` for browsing tools before signup Static metadata at `https://api.mcpanalytics.ai/.well-known/mcp.json` — tool catalog, transport config, auth info. MCP tools: about, discover_tools, tools_schema, tools_run, tools_info, datasets_upload, datasets_list, datasets_read, datasets_download, datasets_update, connectors_list, connectors_query, reports_list, reports_search, reports_view, report_cards, agent_advisor, billing, module_request. ## Comparison vs Alternatives **vs Julius AI:** Julius generates Python code with GPT-4/Claude — different code each run. MCP Analytics uses validated R modules — identical results every time. 1 MCP report (6-16 cards) replaces ~10 Julius prompts. Cost: $1.30-$2.90 vs ~$4.00 on Julius. **vs ChatGPT/Claude:** AI chatbots generate ad-hoc code. No validated methodology, no interactive reports, no PDF export, no R code. MCP Analytics provides structured, reproducible reports. **vs SPSS:** SPSS costs $99/month with a learning curve. MCP Analytics starts at $1.30/report with no software to install. Both use validated methods. MCP adds AI insights and interactive charts. **vs Tableau / DataRobot:** Enterprise platforms at $75-$2,500/month. MCP Analytics offers analytical depth at a fraction of the price for statistical analysis (not dashboard building). ## Audience Pages - [For Business](https://mcpanalytics.ai/for-business): Marketing, revenue, pricing, customer, operations analytics - [For Students](https://mcpanalytics.ai/for-students): Homework, thesis, coursework with R code and citations - [For Analysts](https://mcpanalytics.ai/for-analysts): Validated R analyses with full reproducibility - [For Researchers](https://mcpanalytics.ai/for-researchers): Publication-ready analysis with citations and methodology ## Core Pages - [Home](https://mcpanalytics.ai/) - [How It Works](https://mcpanalytics.ai/how-it-works.html) - [The Engine (MCP)](https://mcpanalytics.ai/the-engine.html) - [Knowledge Layer](https://mcpanalytics.ai/knowledge-layer.html) - [Compare](https://mcpanalytics.ai/compare.html): vs Julius, Tableau, ChatGPT, Akkio, Rows - [Pricing](https://mcpanalytics.ai/pricing.html) - [Marketplace](https://mcpanalytics.ai/marketplace.html): All plans and credit packs - [Sample Reports](https://mcpanalytics.ai/analysis/) - [Free Tools](https://mcpanalytics.ai/free/) - [Your Account + Create Analysis](https://account.mcpanalytics.ai/create-analysis) ## Solution Pages - [Marketing Analytics](https://mcpanalytics.ai/marketing-analytics.html): Campaign ROI, attribution, spend optimization - [Ad Spend Analyzer](https://mcpanalytics.ai/ad-spend-analyzer.html): ROAS efficiency, channel comparison - [Marketing Attribution](https://mcpanalytics.ai/marketing-attribution.html): Multi-touch attribution analysis - [Revenue Forecasting](https://mcpanalytics.ai/revenue-forecasting.html): ARIMA, Prophet, time series forecasts - [Media Mix Modeling](https://mcpanalytics.ai/media-mix-modeling.html): Channel contribution analysis - [Pricing Optimization](https://mcpanalytics.ai/pricing-optimization.html): Price elasticity and sensitivity - [Customer LTV](https://mcpanalytics.ai/customer-ltv.html): Lifetime value with BG/NBD models - [Churn Analysis](https://mcpanalytics.ai/churn-analysis.html): Churn prediction and retention ## E-Commerce Guides - [Shopify Analytics](https://mcpanalytics.ai/shopify-analytics.html): AOV, bundles, churn, demand, geographic, price elasticity - [Stripe Analytics](https://mcpanalytics.ai/stripe-analytics.html): MRR, churn, payment methods, card brand analysis - [Amazon Analytics](https://mcpanalytics.ai/amazon-seller-analytics.html): FBA vs FBM, order value, geographic demand - [Etsy Analytics](https://mcpanalytics.ai/etsy-shop-analytics.html): Listing performance, bundles, SKU tracking - [eBay Analytics](https://mcpanalytics.ai/ebay-seller-analytics.html): Fee analysis, buyer insights, profitability - [Square Analytics](https://mcpanalytics.ai/square-analytics.html): Staff performance, hourly trends, location comparison - [WooCommerce Analytics](https://mcpanalytics.ai/woocommerce-analytics.html): Retention, refunds, product performance ## Learning Resources - [Articles](https://mcpanalytics.ai/articles/): 290+ practical guides to statistical methods and ML techniques - [Tutorials](https://mcpanalytics.ai/tutorials/): 107+ step-by-step analysis walkthroughs - [Blog](https://mcpanalytics.ai/blogs/): 70+ case studies and business insights - [Whitepapers](https://mcpanalytics.ai/whitepapers/): 100+ technical deep-dives ## FAQ **Q: What data formats are supported?** A: CSV files and live connectors (GA4, Google Search Console). Shopify and Stripe connectors planned. **Q: Is my data secure?** A: Encrypted at rest (AES-256-GCM) and in transit (TLS). Isolated per user. Automatic expiration. **Q: Can I reproduce the results?** A: Yes. Every report includes R code and a reproducibility statement. Same data + same module = identical results. **Q: What AI models power the insights?** A: Free and Pay As You Go use Claude Haiku. Pro uses Sonnet. Business uses Sonnet with Opus on request. AI interprets results — it does not generate the analysis. **Q: Is there an API?** A: Yes. MCP protocol for AI agents. REST API also available. **Q: What does a report look like?** A: Browse sample reports at https://mcpanalytics.ai/analysis/ ## Technical Details - Analysis engine: R statistical computing with validated packages - Report format: Self-contained interactive HTML with Plotly.js - AI: Claude (Haiku/Sonnet/Opus) for per-card insights - Infrastructure: Azure VMs, PostgreSQL, Docker containerized R execution - Discovery: 5-signal Reciprocal Rank Fusion - Knowledge layer: 768-dim sentence-transformer embeddings, cosine similarity search - Protocol: Model Context Protocol (MCP) for AI assistant integration ## Optional - [Sitemap](https://mcpanalytics.ai/sitemap.xml) - [llms.txt](https://mcpanalytics.ai/llms.txt): Concise overview - [Robots.txt](https://mcpanalytics.ai/robots.txt)