For AI assistants

Crosstabs MCP Server

Give a stdio-compatible local MCP client the local research loop: import, tab books, governed survey designs, waves, editable report packs, export, and audit history—plus 39 focused statistical tools. Calculations stay deterministic and private project work remains local to the device.

Published by crosstabs.com · Named statistical review pending · Last updated

Install

Codex plugin

codex plugin marketplace add barangaroo/crosstabs-codex-plugin --ref 857a93b32dd22f7e655ec30e3c628f70ba9a14d5
codex plugin add crosstabs@crosstabs

The command pins the verified source SHA for plugin 0.4.0 and crosstabs 1.4.0. Source installation is not native-host review or directory approval. Neither submission, approval nor a public directory listing is claimed.

Python / any local MCP client

pip install crosstabs==1.4.0

# Complete project workflow
crosstabs-headless

# Focused statistical tools
crosstabs

Claude Code

claude mcp add crosstabs-headless -- crosstabs-headless
claude mcp add crosstabs-statistics -- crosstabs

Claude Desktop

{
  "mcpServers": {
    "crosstabs-headless": {
      "command": "crosstabs-headless"
    },
    "crosstabs-statistics": {
      "command": "crosstabs"
    }
  }
}

In claude_desktop_config.json (Settings → Developer → Edit Config).

Cursor / other MCP clients

Add both stdio commands above for full coverage, or connect the aggregate-only remote server at https://mcp.crosstabs.com/mcp.

Verified package; separate source and directory gates

The published crosstabs==1.4.0 passed fresh public-index checks of both local stdio servers: 35 headless tools and 40 statistics-server registrations (39 analytical tools plus runtime status). Both runtime-status calls confirm the executing package version. The exact active Registry 1.4.0 record was verified separately; historical 1.2.2 evidence remains distinct. The plugin source and parity contract is separate evidence from package publication. Native-host review and directory approval remain pending.

The browser worker and local MCP enter the same HeadlessCrosstabsService and share the TabBookProject schema, initial-analysis policy, deterministic statistical kernels, project revision, audit, replay, and undo orchestration. Cross-surface golden tests remain release gates rather than blanket certification. Local persistence is bounded to 8 MiB per serialized project and 32 MiB per database, including audit, undo and replay state; parser import ceilings are not a persistence guarantee. The plugin does not reproduce browser interaction state or optional local connector authorization. The live remote MCP is intentionally narrower: public evidence and explicitly supplied aggregate matrices only, with no project or respondent-row custody.

Plugin source repositoryOfficial MCP Registry 1.4.0 recordPyPI 1.4.0

What the assistant can do with it

Runtime and analysis plans

Executing package/version checks on both servers; portable plans with readable settings, stale-data checks, bounded execution and aggregate review findings

Project workflow

Create, import, profile, inspect, list, export, and audit versioned research projects

Tab-book design

Row sets with governed cases-or-responses denominators for detected multi-select questions, multibanners, filters, preserved weight candidates with fail-closed execution, unweighted table runs, tab books, significance letters, and wave comparisons

Manual coding

Available in the browser workspace, not local MCP: local open-end codebooks, review, approval, materialized variables, and guarded undo

Tracker and reporting

Dataset replacement, schema-drift detection and repair, editable PPTX/DOCX generation, and manual-edit-preserving refresh

Tests

Chi-square (with Yates), G-test, Fisher's exact, McNemar's, CMH stratified test, Breslow-Day, Monte Carlo chi-square

Effect sizes

Cramér's V (bias-corrected), phi, odds ratio with CI, relative risk, risk difference, attributable risk

Measures and planning

Ordinal association, agreement, epidemiology measures, power, sample size, and multiplicity corrections

Diagnostics

Expected counts, residuals, correspondence analysis, sparse-cell warnings, evidence resources, and explicit limitations

Example prompt: “Create a project from this survey, profile cardinality, define age and region banners, run an unweighted tab book, export an editable PPTX with audit evidence. Preserve any weight column as a candidate and do not execute it without typed design authority.”

Source and license

The Python 3.10+ package is declared under the MIT license and is distributed on PyPI as a wheel and downloadable source archive. The development repository is currently private; public GitHub access is not claimed here. The separately packaged plugin source repository is public. Report package or plugin issues to support@crosstabs.com. Core methods share documented conventions with the web calculators, while the package also includes additional methods with their own assumptions and output contracts. See the methods and current review status before relying on results for high-stakes work.

Package contract: crosstabs 1.4.0. Trusted Publisher: GitHub · crosstabs/crosstabs-lite · release-python-mcp.yml · environment pypi.

This release evidence verifies PyPI independently; it does not establish plugin-directory approval or native-host widget review.

Release source SHA: f8bc062b6f76865a9c7629656a933fe98332ab58.Registry verified ; provenance verified .

Prefer plain HTTP?

The free JSON API exposes chi-square, Fisher's exact, odds ratios, and 20+ measures without installing anything.

Frequently asked questions

What is the crosstabs MCP server?
The published MIT-licensed crosstabs 1.4.0 package has two local MCP servers: 39 analytical tools plus runtime status and 35 headless tools, including guarded project handoff, transformations, design-aware methods and saved-table widgets. Fresh public installs of both servers are verified.
Is there an official Crosstabs plugin for Codex?
Crosstabs plugin 0.4.0 is available from a verified public GitHub source revision and pins crosstabs 1.4.0. Source installation is separate from native-host review and directory approval; no directory submission, approval or listing is claimed.
Does the Codex plugin have full Crosstabs workspace parity?
No. Package 1.4.0 covers the local project workflow: projects, imports, filters, typed survey designs, tab books, waves, schema repair, editable report packs, export, and audit history. Manual open-end coding remains a browser-workspace feature, and browser interaction state is not reproduced by the plugin.
Why use an MCP server instead of letting the AI compute statistics itself?
Language models can make arithmetic mistakes. The MCP server executes numerical routines instead; each method still has its own assumptions, and outputs distinguish exact, asymptotic, and simulated inference.
Does the MCP server send my data anywhere?
The local servers run over stdio and keep respondent files and projects on your device. The public endpoint accepts only public evidence identifiers and bounded aggregate frequency matrices; it does not accept respondent rows or retain projects.