Skip to main content

​
Assess Web pages

Use Web assessment to inspect whether agents can discover and use the machine-readable resources published by a public page. It returns evidence and guidance, not a reputation score.

​
Run an assessment

For an agent or automation, request the canonical JSON report:
The human report has four columns:

​
What it checks

​
Discoverability

  • AI crawler directives in robots.txt
  • ARD manifests
  • Agent guides such as llms.txt and declared Markdown alternates
  • Modular guide links, plus pricing.md and auth.md
  • OpenAPI, API catalog, MCP, and A2A discovery documents

​
Usability

  • Final page response, title, and description
  • Retrieval of declared agent resources
  • Markdown responses to Accept: text/markdown
  • Markdown responses to an agent bot user agent

​
Scope and data boundary

Web assessment runs direct HTTP requests from your machine. It follows the requested page to its final URL, then probes only resources on that final origin. It does not render a browser, crawl ordinary links, persist results, reuse a cache, send cookies, authenticate, search the public Web, or invoke an API, MCP server, or WebMCP tool. Missing optional agent resources are advisory. A broken final page, title, or description is a failure. Web assessment inspects a public Web surface. CLI evaluation runs approved scenarios against an existing CLI. MCP evaluation is planned separately.