Skip to content
MCP server

Stats Explorer

Verified statistics tools for assistants: tests, A/B readouts, power analysis and reports

Runs locallyStatus and history
Pricing
Open source
Platforms
CLI, vscode
Runs
Locally
Version
1.5.1
Listing updated
1 day ago

About Stats Explorer

Stats Explorer is an open-source MCP server for analytics and data. It runs locally on your machine and is started by your AI client.

Instead of letting a model guess at numbers, this server hands data analysis to 18 tools that read your CSV, TSV, Excel or Parquet files by column name, so large datasets never go through the chat. Tests cover t-tests, z-tests, ANOVA with Tukey HSD, ANCOVA, MANOVA, chi-square, Fisher's exact, binomial and correlation. A/B and A/B/n readouts report lift with intervals, sample ratio mismatch checks, Holm-adjusted comparisons and Bayesian probability to beat the control, and there is sample-size planning.

Each result includes an effect size, confidence interval, a plain-English interpretation and warnings when an assumption looks weak. 43 reference cases are checked against SciPy and statsmodels on every commit. Columns, groups and report sections the agent proposes are shown to you for confirmation before anything runs.

It runs locally over stdio with no network calls or API key. Optionally, TypeSafe's Jev model can independently check the test plan and the assistant's written conclusions.

Install

Run in a terminal

claude mcp add stats-explorer-mcp --env TYPESAFE_API_KEY='<TYPESAFE_API_KEY>' --env STATS_EXPLORER_APPROVAL_SECRET='<STATS_EXPLORER_APPROVAL_SECRET>' -- npx -y stats-explorer-mcp

Add to claude_desktop_config.json

{
  "mcpServers": {
    "stats-explorer-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "stats-explorer-mcp"
      ],
      "env": {
        "TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>",
        "STATS_EXPLORER_APPROVAL_SECRET": "<STATS_EXPLORER_APPROVAL_SECRET>"
      }
    }
  }
}

Add to ~/.cursor/mcp.json

{
  "mcpServers": {
    "stats-explorer-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "stats-explorer-mcp"
      ],
      "env": {
        "TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>",
        "STATS_EXPLORER_APPROVAL_SECRET": "<STATS_EXPLORER_APPROVAL_SECRET>"
      }
    }
  }
}

Add to .vscode/mcp.json

{
  "servers": {
    "stats-explorer-mcp": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "stats-explorer-mcp"
      ],
      "env": {
        "TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>",
        "STATS_EXPLORER_APPROVAL_SECRET": "<STATS_EXPLORER_APPROVAL_SECRET>"
      }
    }
  }
}

Add to ~/.codeium/windsurf/mcp_config.json

{
  "mcpServers": {
    "stats-explorer-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "stats-explorer-mcp"
      ],
      "env": {
        "TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>",
        "STATS_EXPLORER_APPROVAL_SECRET": "<STATS_EXPLORER_APPROVAL_SECRET>"
      }
    }
  }
}

Add to your client's MCP settings (most use this shape)

{
  "mcpServers": {
    "stats-explorer-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "stats-explorer-mcp"
      ],
      "env": {
        "TYPESAFE_API_KEY": "<TYPESAFE_API_KEY>",
        "STATS_EXPLORER_APPROVAL_SECRET": "<STATS_EXPLORER_APPROVAL_SECRET>"
      }
    }
  }
}

Configuration

  • STATS_EXPLORER_ROOTSFolders the tools may read data files from and write reports to (separated by : on macOS/Linux, ; on Windows). Defaults to the client's workspace roots, or the current directory.
  • STATS_EXPLORER_STREAM_MBCSV/TSV/Parquet files above this size (MB) are streamed in parallel passes; any size works. Default 512.
  • STATS_EXPLORER_WORKERSParallel readers for large files. Default: all CPU cores but one (max 16).
  • STATS_EXPLORER_CACHEKeep columns read from large files for fast follow-up questions: ask (default), on or off.
  • STATS_EXPLORER_CACHE_DIRFolder for saved large-file columns. Default: the system temp folder.
  • STATS_EXPLORER_MAX_FILE_MBRefuse data files larger than this (MB). Default: no limit.
  • STATS_EXPLORER_GUIDEDon (default): classify each request and follow the six-step workflow for hypothesis tests. off: expose the individual test tools directly.
  • STATS_EXPLORER_CONFIRM_SELECTIONSon (default): the user confirms every column, chart type and report section before anything runs. off: run with the AI's choices.
  • STATS_EXPLORER_CONFIRM_VIAauto (default): forms where the client supports them. chat: always ask in the conversation (recommended for Claude Code).
  • STATS_EXPLORER_HIDE_LINKSet to 1 to remove the one-line Stats Explorer app link from results.
  • STATS_EXPLORER_JEVIndependent checks by Jev (TypeSafe): off (default), cooperate (the AI calls your Jev MCP server), enforce (Stats Explorer calls the Jev API) or orchestrated (signed decisions from an orchestrator).
  • STATS_EXPLORER_JEV_CHECKPOINTSComma list of plan, conclusions, data. Default plan,conclusions (data sends column names and text samples to Jev).
  • TYPESAFE_API_KEYEnforce mode only: TypeSafe API key Stats Explorer uses to call Jev.
  • STATS_EXPLORER_APPROVAL_SECRETOrchestrated mode only: shared secret (32+ characters) the orchestrator signs decisions with.
  • STATS_EXPLORER_AUDIT_LOGPath of a JSON Lines file recording every Jev checkpoint decision.

What it does

Statistical tests
t-tests, ANOVA with Tukey HSD, chi-square, Fisher's exact and more.
A/B readouts
Lift, sample ratio checks and Bayesian win probability.
File-based analysis
Read CSV, TSV, Excel and Parquet without pasting data.
Guided workflow
Hypothesis tests follow a six-step guide with confirmations.
Plain-English results
Effect sizes, intervals and assumption warnings.
Optional Jev check
A second model reviews plans and conclusions.

Good for

  • Analyse an A/B test result from a CSV
  • Find which groups differ after ANOVA
  • Plan the sample size for an experiment
  • Rank which columns drive an outcome

Pricing

Open source

Free; runs locally

Stats Explorer FAQ

Does it send my data anywhere?

No, it runs locally and makes no network calls unless Jev's enforce mode is on.

How accurate is it?

43 reference cases are checked against SciPy and statsmodels to six significant figures.

Which file types work?

CSV, TSV, Excel and Parquet.

Can I use it in VS Code?

Yes, via an extension for Copilot's agent mode.

Is Stats Explorer free?

Yes. Stats Explorer is open source, so you can use and self-host it for free.

How do I add Stats Explorer to Claude Code?

Run this command in your terminal: claude mcp add stats-explorer-mcp --env TYPESAFE_API_KEY='<TYPESAFE_API_KEY>' --env STATS_EXPLORER_APPROVAL_SECRET='<STATS_EXPLORER_APPROVAL_SECRET>' -- npx -y stats-explorer-mcp

What are the best alternatives to Stats Explorer?

The closest alternatives to Stats Explorer are REA, Codebase Memory, GitHub, Serena MCP: the IDE for your agent and XERJ.

Alternatives to Stats Explorer

  1. REA

    MCP server

    CLI and MCP server that lets coding agents reverse engineer apps and binaries

    Runs locally
    • Open source
    • 464K GitHub stars
    • Developer tools
  2. Codebase Memory

    MCP server

    Local MCP server that builds a code knowledge graph for AI coding agents

    Runs locally
    • Open source
    • 48.7K GitHub stars
    • Knowledge and memory
  3. GitHub

    MCP server

    Official MCP server connecting AI assistants to GitHub repos, issues and workflows

    Not checked yet
    • Remote
    • Open source
    • 34.5K GitHub stars
    • Developer tools
  4. MCP toolkit giving coding agents symbol-level code retrieval and editing

    Runs locally
    • Open source
    • 31.5K GitHub stars
    • Developer tools
  5. XERJ

    MCP server

    Local search engine for AI agents with cited answers over your code and docs

    Runs locally
    • Open source
    • 19.2K GitHub stars
    • Knowledge and memory
  6. Open-source analytics app where an AI agent builds charts and dashboards from your data

    • Remote
    • Open source
    • 7.7K GitHub stars
    • Analytics and data