Glossary

Definitions of the terms used throughout this documentation and in audit reports.

AI-readiness: How well a large language model or autonomous agent can read a documentation site and call its API correctly without guessing, measured by the 30-criteria rubric.
Audit: A single evaluation of one documentation site against the rubric, producing a deterministic JSON report scored from 0 to 100 with per-criterion HTTP evidence.
Base URL: The root URL of the documentation site being audited, such as https://docs.stripe.com; all host-root discovery files are fetched relative to it.
Criterion: One of the 30 individual checks (identified A1 through F4) that make up the rubric; each contributes a fixed number of points to the total score.
Category: One of the six groups of criteria — Discovery, Page artifacts, API spec, Content, Hygiene, and Agent Surface — whose points sum to the 0–100 total.
Run: A queued or completed execution of an audit, identified by a UUID run_id that you poll until its status becomes done.
llms.txt: A standardized plain-text file at a site's root that summarizes the site for large language models and links to its most important pages and resources.
MCP server: A Model Context Protocol endpoint that lets an agent client such as Claude discover and call the audit tools over an authenticated transport.
WebMCP: Declarative markup on an HTML page that exposes a form as an agent-callable tool, so a browsing agent can invoke it without bespoke integration.

Terms that map to JSON fields in a report (total_score, categories, criteria, status, evidence_url) are documented inline on the Get an audit reference page.