On-demand audits
Ask your MCP client to call a tool with these example inputs. Replace example properties and URLs with your authorized ones. Inputs shown as optional by the schema may be required when starting a new session.
Page audit
gsc_page_audit accepts 1–5 URLs in one property and an optional period. It combines current HTTP HTML with exact-page Search Console totals and observed queries. Text matches are literal, not semantic judgments.
gsc_page_audit · MCP input
{
"siteUrl": "https://example.com/",
"urls": [
"https://example.com/article"
],
"preset": "last_28_days"
}Indexing audit
gsc_indexing_audit accepts either 1–10,000 explicit URLs or 1–100 sitemap roots, not both. If neither is provided it uses registered sitemaps. Selection is sitemap_order or recent_lastmod; inspectLimit is 1–10 per call, default 10. This is sampled indexed-state evidence, not a live test or indexing request.
gsc_indexing_audit · MCP input
{
"siteUrl": "https://example.com/",
"sitemapUrls": [
"https://example.com/sitemap.xml"
],
"selection": "recent_lastmod",
"inspectLimit": 5
}Internal links
gsc_internal_link_audit starts with 1–100 startUrls, optionally using up to 100 sitemaps. maxPages defaults to 100 (maximum 500); batchSize defaults to 20 (maximum 50). Incoming links are observed, not proof that other links or pages do not exist. Redirect aliases share their effective HTTP source; requested URLs and redirects remain evidence. Suggestions are literal text candidates.
gsc_internal_link_audit · MCP input
{
"siteUrl": "https://example.com/",
"startUrls": [
"https://example.com/"
],
"maxPages": 100,
"batchSize": 20
}Change impact
gsc_change_impact compares observed metrics before and after a change. A transient analysis needs no saved history. Its observations and optional controls do not establish causation. See the history guide for records and snapshots.
gsc_change_impact · MCP input
{
"siteUrl": "https://example.com/",
"action": "analyze",
"change": {
"date": "2026-01-15",
"urls": [
"https://example.com/article"
]
}
}Migration audit
gsc_migration_audit requires authorized source and target properties, a migration date and 1–500 explicit oldUrl/newUrl mappings. Batch size defaults to 2 (maximum 5). Many old URLs may share one destination; destination metrics are not summed per mapping. HTTP is current while Google index evidence may lag. Redirects do not prove migration success.
gsc_migration_audit · MCP input
{
"sourceSiteUrl": "https://old.example.com/",
"targetSiteUrl": "https://example.com/",
"migrationDate": "2026-01-15",
"mappings": [
{
"oldUrl": "https://old.example.com/article",
"newUrl": "https://example.com/article"
}
]
}Limits and coverage
Each call has up to 60 seconds and 12 Google attempts; website resources have a 15-second timeout, 2 MiB per HTML page and up to five redirects. Sitemap discovery handles up to 10 files per call, 50 MiB accumulated uncompressed data and 10,000 URLs. A sitemap with more than 10,000 entries needs smaller child maps or explicit URLs. Automatic traversal respects robots.txt, including redirects. Only received HTTP HTML is parsed: JavaScript-rendered content and links are unknown.
Continue and recover evidence
When data.nextCursor is not null, call the same tool again with the returned value as input cursor, retaining the original parameters unchanged, or provide the property identifier and cursor only. Migration requires both property identifiers. Sessions are isolated by profile, properties and allowlist, expire after 15 idle minutes or restart, and can be evicted for capacity. No work continues after a response; there is no scheduler or background monitoring. Replace the placeholder below with the returned cursor.
gsc_indexing_audit · MCP input
{
"siteUrl": "https://example.com/",
"cursor": "replace-with-data-nextCursor"
}complete means the bounded work finished, not that discovery or the graph is exhaustive. Check status, warnings, failures and coverage. For full evidence use meta.datasetId with gsc_get_dataset, following its own cursor. Group rows by index, order by part, concatenate evidence and parse the resulting JSON. The summary has a separate evidence index. No additional Google request is needed.