Clutter MCP server

clutter-mcp — synthetic data & synthetic document generator (MCP server)

A Model Context Protocol stdio server that lets an AI agent drive the whole Clutter pipeline — invent a believable synthetic company, mass-produce the documents, spreadsheets, emails, images and datasets it would really have, poll for completion, and fetch download URLs — through Clutter's public REST API, authenticated with an API key.

Use it to fill dev/test/demo systems (SharePoint, CRMs, file shares) with realistic content — synthetic documents and synthetic data that stand in for the real thing. Generate test data, demo data and training data, or test documents and training documents, on demand — or give an AI agent believable data to reason over, all without touching real or sensitive data.

It's a thin HTTPS client with no dependency on the rest of the Clutter codebase.

Quick start

(cltlive…, shown once).

1. Create a free account at https://clutter.run, open Settings, and mint an API key

1. Add the server to your MCP client config (Claude Code, Claude Desktop, etc.):

{
  "mcpServers": {
    "clutter": {
      "command": "npx",
      "args": ["-y", "clutter-mcp"],
      "env": {
        "CLUTTER_API_KEY": "clt_live_…"
      }
    }
  }
}

That's it — CLUTTERAPIURL defaults to https://clutter.run/api, so only the key is required.

Configuration

Run directly

CLUTTER_API_KEY=clt_live_… npx -y clutter-mcp

Typical agent flow

auto-create a project). Free.

1. build_org — describe a company in a sentence → returns an orgId (omit projectId to

1. waitfororg — block until the company build is ready.
2. create_run — generate content against the company:

• doc_generator → a batch of documents (docx/pdf/xlsx/eml/jpg), folder-organised.
• data_generator → one tabular dataset (xlsx/csv/json) with an exact row count.
• docmetadatagen → one metadata record per document of a prior doc run.

single ZIP (folder tree preserved — ready to drop into SharePoint or a file share).

1. waitforrun — block until the run is complete.
2. listrundocuments + getdocumenturl, or buildzip + getzip_url for the whole run as a

Tools

getdocumenturl, buildzip, getzipurl, deleterun

• Identity / usage: clutterwhoami, getusage
• Projects: listprojects, createproject, delete_project
• Companies: listorgs, buildorg, getorg, waitfororg, queryorg, delete_org
• Runs: listruns, createrun, getrun, waitforrun, listrun_documents,

buildorg and createrun are asynchronous (return an id immediately); use the waitfor tools to block until a terminal state, or poll getorg / getrun. Each tool maps to a REST endpoint and returns the raw JSON response.

Billing

Company builds, metadata and "ask the company" are free. New accounts get 10 free documents + 100 free data rows, then pay-as-you-go (all prices USD: $0.10/document, $0.06/10 data rows). A 402 from create_run means insufficient credit — top up at https://clutter.run/billing.

Loading content into SharePoint

Shortened. The full README is on GitHub.