Improve search performance

Munin's analytics answer the post-click half of a traffic question: which pages got viewed, where visitors came from, what they searched for on the site and found nothing. A connected search engine answers the pre-click half: what people typed, how often the site was shown, and where it ranked. Neither half is actionable alone. Together they point at a specific page and a specific gap, and the same agent can then fix it.

The connector is read-live: nothing from the search engine is stored in Munin, so every call reflects what the engine holds right now.

Tools

seo_list_properties — the verified sites this connection can report on. Call it first if you don't know the siteUrl.
seo_list_queries — queries by impressions, with clicks, ctr and avgPosition.
seo_list_pages — the same measures per page URL.
seo_inspect_url — one URL's index record: indexed or not, HTTP status at last crawl, when it was last crawled and first discovered.
seo_submit_urls — ask the engine to (re)crawl up to 500 URLs under the property.

connectionId is only needed when the org has several active seo connections. siteUrl is only needed when the account has several verified properties — with one, it resolves itself.

Two engines, one shape

Bing Webmaster Tools and Google Search Console both sit behind these tools and return the same normalized fields, so the loop below reads the same either way. Where they differ:

Only Bing accepts seo_submit_urls. Google Search Console has no URL-submission endpoint (its Indexing API covers only job postings and broadcast events), so the tool refuses for a Google connection and says so. Ask for a recrawl on Bing; for Google, publishing and an updated sitemap are what you have.
seo_inspect_url fills different fields per engine. Bing gives httpStatus, discoveredAt and inboundAnchorCount; Google gives detail — its coverage state, the most useful single string here ("Submitted and indexed", "Crawled - currently not indexed", "Discovered - currently not indexed"). Report the fields you got; a null means the engine doesn't expose it, not that the value is zero.
Bing's index reaches Copilot and part of ChatGPT search, so it is not only legacy SEO. Connecting both engines is normal, and an org can have one connection per engine.

Read the numbers correctly

Three properties of this data will mislead you if you treat it like Munin's own analytics:

It lags. Reporting is 2–3 days behind on both engines. A page published yesterday has no data yet, and that is not a problem to diagnose.
The window is what was covered, not what you asked for. Bing aggregates into whole weeks ending Friday, and either engine simply has no rows for days it holds no data. So quote window from the result, never the from/to you passed. When it is null, nothing fell in range at all.
avgPosition is impression-weighted and can be null. Null means the engine reported no position for that row, not position zero. A query with 4 impressions and position 3 is noise; sort your attention by impressions.

The loop

Find the gap. seo_list_queries over a wide window. The interesting rows are high impressions with low clicks — the engine is showing the site for that query and nobody is choosing it. Very high avgPosition (a large number) with real impressions is the other shape: the query matters but the site ranks far down.
Confirm it from the other side. analytics_list_zero_result_searches shows what visitors searched for on the site and got nothing — a query appearing in both lists is a content gap confirmed twice. analytics_list_referrer_hosts and analytics_list_traffic_sources tell you whether search traffic is actually arriving.
Find the page that should answer. kb_search and cms_search_entries with the query wording. Three outcomes worth distinguishing: nothing exists (write it), something exists but doesn't use the customer's words (rewrite it), or the right page exists and ranks badly for reasons content can't fix (report that, don't churn the page).
Check the page is even indexed before rewriting it — seo_inspect_url. A page the engine has never crawled, or last crawled with a non-200 status, has an infrastructure problem; editing the copy will not move it.
Fix it. cms_update_entry for marketing pages, kb_create_document or kb_update_document for support content. Use the customer's phrasing from step 1 in the title and opening paragraph.
Submit it, on a Bing connection. seo_submit_urls with the URLs you changed. Submissions spend a per-site daily quota, so submit what changed and nothing else — never the whole sitemap "to be safe". The result reports the quota left; if the call is rejected for exceeding it, the batch was not submitted at all and there is nothing to undo. On a Google connection this step does not exist; say so rather than looking for another tool.
Come back later, not now. Effects need a recrawl plus the reporting lag. Re-checking the same query an hour later tells you nothing.

What this cannot do

It cannot see queries with very few impressions — search engines withhold low-volume and personally identifying queries. An absent query is not proof of zero impressions.
It cannot attribute revenue. Clicks here and conversions in analytics_get_funnel are separate datasets over different identity models; describe them side by side rather than dividing one by the other.
Submitting a URL is a request, not a guarantee: the engine decides whether and when to crawl, and indexing is never promised. Report what was submitted, not what will rank.

skill://analytics/track-website-traffic — the post-click half, from Munin's own tracker: which pages were viewed, from which referrers, and what visitors searched for on the site.
skill://connectors/connect-external-system — how a connection is created and credentialed in the first place, including the OAuth redirect that Google Search Console needs.
skill://cms/publish-entry and skill://kb/review-content — where step 5 of the loop actually happens.