Integrations

Use the public definitions directly, or keep a versioned copy in your publishing tools.

The vocabulary and marks are CC0. Website prose is CC BY 4.0; code is MIT. Requests to hosted endpoints reach the hosting provider. Inline SVG and downloaded copies work independently.

Integrations

Static endpoints

/levels.json: The canonical vocabulary: definitions, decision procedure, edge cases and suggested mappings. Localised copies are available at /<lang>/levels.json.
/spec.schema.json: JSON Schema for the vocabulary file itself.
/levels.schema.json: JSON Schema for a declaration: level 0–5, optional surfaces, translation and review information.
/ecosystem.json: A directory of related standards and projects; listing does not imply endorsement.
/llms.txt: A short text index for tools that choose to read this convention. Google does not require it.
/llms-full.txt: The full English vocabulary as plain text, including its limitations.
/feed.xml: A release feed. Pin a release when reproducibility matters.
/sitemap-index.xml: Canonical pages and the translations that are actually available.

Local MCP server

The repository includes a stdio server with four tools: classify, get_level, list_levels and get_spec, plus a usagescale://levels.json resource. Install its dependencies once; the server reads local data and makes no runtime network requests.

As checked on 10 October 2026, usagescale-mcp is not published in the public npm registry. Install from the repository:

Install and run from source

git clone https://github.com/karauda/ai-usage-scale.git
cd ai-usage-scale/mcp
npm ci
node server.js

MCP client configuration

Replace the absolute path with your checkout location.

{
  "mcpServers": {
    "usagescale": {
      "command": "node",
      "args": ["/absolute/path/to/ai-usage-scale/mcp/server.js"]
    }
  }
}

The classifier returns a decision-tree suggestion. Verify the full definition, especially Level 3, which requires meaningful human review even though the current tree stops before asking about it.

Declare the work

Each level provides a sentence, an SVG mark and metadata. The HTML metadata is a project convention. JSON-LD uses an explicitly namespaced extension; aiUsageLevel is not a registered Schema.org property.

There is no dedicated Google Search feature for this scale. Do not use a declaration as a quality score, compliance certificate or evidence of permission to train a model.

A publisher may choose /.well-known/ai-disclosure.json as a discovery location. That path is a project proposal, not a registered well-known URI. This website currently publishes no numeric self-declaration: the revised content awaits human editorial review.