Menu
The platform
The agent layer documented in the quickstart is one face of SiteTR. Behind it sits a platform that runs the whole job — this page is the map, and every property below has its own page.
What SiteTR runs
The home page sells six disciplines; this section documents every property that delivers them — plus the surfaces you work through. Each reads the same stored evidence, so a rank drop, a decay verdict, and an audit finding can be the same story told three ways — SiteTR keeps them joined.
Finding → verified fix
Most tools end at the dashboard. SiteTR’s loop runs all the way to verified: a finding becomes a recommendation, the recommendation becomes an intervention — a patch, a GitHub pull request, a manual change, or a social post — optionally routed through customer approval, and every applied change gets a verification pass that re-checks the live page. Outcomes land in the ledger so the next report can say what actually moved.
Where you work
Your AI agent. The front door — every property above is reachable as an MCP tool (see the tool reference). Ask a question, get an answer with its data source named.
The REST API. The same engine the panels use — /api/audit, /api/google/*, /api/history/* and friends, session login, JSON everywhere.
The panels. Agency panel, customer portal, and admin — the team surfaces for reports, approvals, watch lists, and Search Console setup. Covered in Panels.
What it reads
First-party crawls (raw HTML and rendered captures), Google Search Console and GA4 via OAuth, Chrome UX Report and PageSpeed Insights, live SERPs, and Wikidata and friends for entity resolution. Every number carries provenance: which provider, fetched when. When the honest answer is “not enough data”, the engine says INSUFFICIENT_DATA instead of fabricating one — see Evidence & verify.