M
MCP Directory CURATED DIRECTORY
911 MCP servers indexed
中文
← Back to directory
M

MCP Analytics

Community
The statistical analyst in your AI chat — bring data and a question, own a citable, re-runnable analysis.
How to install ↓ Compare GitHub source repository ↗
Category
Other #124 of 230
Stars
★ 8 Popular
Transport
stdio (local process) · Streamable HTTP
Runtime
Node.js 18+ · Hosted (no local runtime)
Credentials
API key / credential required
License
MIT
Last commit
Tools
18
53FMRS · D

MCP Analytics extends AI assistants with citable, reproducible, and methodologically validated statistical reports. Its tiered model from free Snapshots to deep Deck studies is careful to bill only on success. It offers multiple transport options, auth-free discovery, and a browsable tool catalog, making it easy to try. Main caveats: commercial nature, cloud processing (not for sensitive data), and it's a third-party service, not official.

Strongest · Setup experience 13/20 Weakest · Reliability 8/20

Reliability
8/20
Security and permissions
12/20
Maintenance
8/20
Documentation
12/20
Setup experience
13/20
Why each score
Reliability 8/20
The repo explicitly states it is 'the public listing and documentation repository'; the API server code is not in source, so static review cannot verify declared tool behavior, and no committed tests or CI exist. Core functionality depends on a remote cloud build pipeline (30–45 min queues) with error handling outside local control. Score 8: the discovery endpoint offers copyable curl verification, but key paths are not statically verifiable.
Security and permissions 12/20
Credentials support OAuth2/PKCE and API keys over HTTPS; isolated containers and encryption are claimed. No red lines found (no malware, no leaked tokens, dangerous operations gated by paid confirmation). Deductions: the server is closed-source, data flows to a third-party cloud processor, and the 'no data retention' claim is a trust assertion with no reviewable evidence; deletion and retention practices are promises only.
Maintenance 8/20
Version 1.0.7, 6 open issues, MIT license in manifest — though the README copyright section says 'All Rights Reserved', contradicting the MIT badge. Support email and security doc link exist. No visible release cadence, dependency update, or security-response process evidence; active but governance/versioning gaps, so 8.
Documentation 12/20
Well-layered docs: quickstart, per-client config examples (Claude/Cursor/Windsurf/Claude Code), tool catalog, tier/pricing pages, troubleshooting checklist, anonymous discovery examples. Deductions: tool parameter schemas are not in the repo (online-only), key claims like 'ephemeral, no persistence' lack reviewable evidence, and performance statements are vague ('Handles large datasets').
Setup experience 13/20
Setup path is very clear: one-line npx config, direct HTTP, and zero-config OAuth, each with complete JSON examples, plus an account-free way to browse tools. However, within the 15-point ceiling there is no in-repo CI+tests execution evidence and the npm package contents cannot be statically verified, so 13.

Static review · not runListed 2026-09-08

Read the FMRS scoring method →

Fit and risk

What it can accessRuns commands or codeUses the network

Best for

  • Business analysts and data science enthusiasts seeking rigorous statistical analysis through an AI chat interface.
  • Teams needing citable, reproducible analyses with methodology descriptions for papers, presentations, or compliance.
  • Users who want to own their analysis modules and rerun them on fresh data without paying again.
  • Adopters of Claude Desktop, Cursor, Windsurf, or other MCP clients who want to analyze data directly in their workflow.

Not for

  • Users needing a free or open-source solution (service is commercial, credit-based).
  • Users who just need simple SQL queries or file system operations (more specialized, free MCP servers exist).
  • Users requiring fully local processing to avoid cloud (data is sent to API for processing).
  • Users needing very large-scale data processing or non-statistical custom coding (platform focuses on statistical methods).

Required permissions

  • Tools can read uploaded CSVs or pull data from connected sources (GA4, GSC).
  • Running analyses involves calling remote cloud APIs; network access required.
  • Auth via OAuth2 (Auth0) or API key.
  • Billing tool can affect credit usage within this API (not external payments).
  • May create or update remote connector data if supported.

Risks and side effects

  • Data is processed in the cloud, so sensitive information could be exposed despite encryption and ephemeral processing.
  • Commercial service relies on credits; builds that fail are not billed, but continued use requires payment.
  • Non-official, third-party service may have rate limits or downtime.
  • Potential for errors in AI-generated analyses if used without scrutiny; platform claims verification, but users should review.

Setup

Before you start

Runtime：Node.js 18+ · Hosted (no local runtime)

MCP_ANALYTICS_API_KEY requiredsecret API key for MCP Analytics, starts with mcp_; get it from account settings at account.mcpanalytics.ai.
  1. Sign up at account.mcpanalytics.ai and copy your API key (starts with mcp_). You receive 500 welcome credits — no credit card required.
  2. Add the mcpServers configuration to your MCP client (Claude Desktop, Cursor, Windsurf, etc.).
  3. Alternatively, connect via Streamable HTTP using the token URL with your API key, or via OAuth2 for zero-config.
  4. Restart the client and start asking questions.
claude_desktop_config.json 
{
  "mcpServers": {
    "mcpanalytics": {
      "command": "npx",
      "args": [
        "-y",
        "@mcp-analytics/mcp-analytics"
      ],
      "env": {
        "MCP_ANALYTICS_API_KEY": "mcp_your_key_here"
      }
    }
  }
}

Shown for Claude Desktop. Other clients may use a different file or key (VS Code uses "servers") — the configurator below converts it.

.vscode/mcp.json
{
  "servers": {
    "mcpanalytics": {
      "command": "npx",
      "args": [
        "-y",
        "@mcp-analytics/mcp-analytics"
      ],
      "env": {
        "MCP_ANALYTICS_API_KEY": "mcp_your_key_here"
      }
    }
  }
}

Goes in your project's .vscode/mcp.json (VS Code uses a "servers" key).

Terminal
claude mcp add mcpanalytics -e MCP_ANALYTICS_API_KEY=mcp_your_key_here -- npx -y @mcp-analytics/mcp-analytics

Run it in a terminal; replace any <…> placeholders with your own values first.

Generate and validate client config

Check that it works

Restart your MCP client and confirm tools like create_analysis, discover_tools, and datasets_upload appear; alternatively run npx -y @mcp-analytics/mcp-analytics and look for the '[mcp-analytics] Connected' line.

Troubleshooting

  1. Ensure your config file is valid JSON.
  2. Restart the MCP client completely.
  3. Verify your API key starts with mcp_.
  4. Check the client's developer console for errors.
  5. Run the npx command in a terminal manually to se errors (should output 'Connected to ...').
  6. Contact support at [email protected].

Things to try

Once connected, you can ask your AI assistant things like:

  • Upload sales.csv and find what drives revenue
  • Commission a deck-tier churn analysis on this dataset
  • Re-run my forecasting analysis on fresh GA4 connector data
  • What statistical test should I use for this survey data?

Tools 18

create_analysis writes
Commission a new analysis from a plain-language question at your chosen tier (snapshot, JSON, brief, or deck).
build_status read-only
Track a build: stage progress, queue position, and report link.
run_analysis writes
Run an analysis you own (or one discovered via discover_tools) on fresh data.
modify_analysis writes
Turn an existing analysis into a new version — reword the question or change framing.
discover_tools read-only
Browse what you can run: your commissioned analyses plus the prebuilt library.
tools_schema read-only
Get an analysis's parameter schema — always call this before run_analysis.
datasets_upload writes
Secure data upload with encryption.
datasets_list read-only
List and search your uploaded datasets.
Show 10 more tools
connectors_list read-only
List available data source connections (e.g., GA4, GSC).
connectors_query read-only
Pull live data from a connected source.
reports_view read-only
Get a shareable browser link for a report.
reports_list read-only
Your report library — searchable in plain language.
report_cards read-only
Browse a delivered report's individual cards (charts, tables, insights).
ask_library read-only
Ask one question across all delivered analyses; get a synthesized answer with citations.
agent_advisor read-only
AI help desk — which analysis fits your question and how to read results.
billing read-only
Usage and credit management.
account_link read-only
Get a link to account pages for anything not doable in chat.
about read-only
Platform documentation and info on how it works, tiers, usage.

Use cases

Upload a CSV and ask a question (e.g., 'Does ad spend drive sales?') and get a statistical answer.
Segment your customers (clustering) or predict churn on your data.
Forecast next quarter's revenue from time series data.
Run hypothesis tests or regression on survey data.
Share citable, reproducible reports in papers, decks, or regulatory filings (APA/MLA/BibTeX).

Supported clients

Claude Desktop
Cursor
Windsurf
Claude Code

Listed from the project's documentation, not tested by this site.

Overview

MCP Analytics is a statistical analyst inside your AI chat. You provide data (CSV or live connectors) and a question; a team of specialist agents builds a custom analysis specific to your data, validates methodology, and delivers an interactive, citable report. The analysis is yours—it lives in your private library, reruns on fresh data at a fraction of the cost, and is queryable from Claude, Cursor, or any MCP client. Multiple depth tiers are available, from instant Snapshots to full Deck studies, with pre-built cornerstone modules covering t-tests, regression, clustering, time series, and more. The platform prioritizes rigorous statistical methods over decorative charts.

Similar servers

BioMCP 75 · B

One binary. One grammar. Evidence from the biomedical sources you already trust.

★ 659 · Tools 11 Compare with this →
Compare with other MCP servers

Source revision 33a358038109 Data synced 2026-10-19 Read the FMRS scoring method