Skip to content

analyze_page

Analyze one absolute HTTP or HTTPS URL.

Parameters

Parameter Type Required Default
url string Yes -
strategy string No both
categories string[] No performance, SEO, accessibility, best practices
locale string No PSI default

Valid strategies are mobile, desktop, and both.

Valid categories are:

  • performance
  • seo
  • accessibility
  • best-practices
  • agentic-browsing (experimental)

Response

The response contains results and errors. Every successful result has:

  • metadata: input strategy, PSI timestamp, Lighthouse version, redirects, warnings, and runtime errors.
  • fieldData: page and origin CrUX measurements. Missing data remains absent; page-to-origin fallback is explicit.
  • labData: open category map, lab metrics, Lighthouse 13 insights, diagnostics, audit details, metric savings, and entity classifications.
  • rawResponse: the complete successful Google API JSON response, including unknown fields and details of audits omitted from the normalized summary.

Field metrics use the upstream p75 rating and preserve histogram distributions. Lab metrics retain their Lighthouse score and unit instead of receiving field-data ratings.

Use rawResponse.lighthouseResult.audits["lcp-breakdown-insight"].details for Google's LCP phase breakdown when supplied, even if Lighthouse marked the audit as passed. Other returned audits, configuration, environment, network details, timings, and screenshot data remain accessible in the same raw report. The server does not remove fields based on scores or a fixed allow-list.

rawResponse is included by default and requires no additional API request. It preserves JSON values and types, not byte-for-byte formatting. Reports may be large or contain sensitive page data; the MCP client may also have its own output limits. Fields or trace artifacts Google does not return are not created.

Examples

Analyze https://www.devleader.ca on mobile. Compare real-user LCP, CLS, and INP
with the Lighthouse lab metrics and prioritize the current insights.
Run the experimental agentic-browsing category for https://www.devleader.ca
and summarize its WebMCP and llms.txt findings.