Catch drops in your site's positions and queries in chat

The agent calls the fourth version of the Webmaster API: first it gets the user identifier, then the list of your hosts, and from there it goes through them section by section — search queries with impressions, clicks and average positions, URLs crawled by the robot with their server responses, pages in search, external links, sitemaps, diagnostics problems and SQI history. It compares periods in pairs so that a drop is visible by query and by page.

The main Webmaster trap is what counts as a site. It is not a domain but a host: example.com, www.example.com and the https version are treated as different sites, each with its own rights and its own statistics. The agent works with the host verified under your login and never mixes their data. It assembles the picture host by host, so comparing a domain with its www version is two passes rather than one request, and a gap between their numbers means different sites, not a broken export.

The agent writes within limits and to the point: it adds a site to the list, starts a verification check, adds a missing sitemap and submits updated URLs for recrawl. Before submitting it looks at how much of the daily recrawl quota is left, so it is not burned on service pages, and afterwards it follows the status of the queued tasks. None of this changes the site's content, its positions or the search results: the robot simply arrives sooner.

The connection is a sign-in to a Yandex account with a code, carrying host owner and verification rights; no tokens to copy, and access renews itself. The agent polls query analytics noticeably slower than the other sections — that endpoint is limited more tightly — so a report on a large site is assembled in parts and it tells you which period is already counted. Long ranges are taken in windows and stitched together, and when data is missing because of rights or an unverified host, it says so directly.