Events Analytics

Track what visitors do on your docs — not just what they view.

Tracked Events#

Event	What it means
AI chat opens	Visitor opened the AI assistant panel
AI queries	Number of questions asked to the AI
Code copies	Visitor copied a code block
Outbound link clicks	Visitor clicked an external link
Sidebar navigation	Visitor navigated via the left sidebar
Language switches	Visitor changed the display language

How to Open#

Float Widget → Analytics tab → scroll to the Events section.

Events are shown as a breakdown with counts and percentages for the selected time range.

How to Use This Data#

High AI opens, low queries → Visitors are curious about AI but find the suggested questions unclear. Try updating your custom questions.

High code copies → Your code examples are useful. Consider adding more.

High language switches → Visitors want content in their language. Enable more translations →

High outbound clicks on a specific page → That page is a good traffic handoff point to your product.

Drill into a single visitor (MCP, PRO+)#

You can investigate what one specific anonymous visitor actually did, end-to-end, via the MCP server.

get_top_visitors(period: "7d", limit: 25)
  → [{ visitor_id: "a1b2…", pageview_count: 14, first_seen, last_seen, country }, …]
 
get_visitor_activity(visitor_id: "a1b2…", period: "7d")
  → { first_seen, last_seen, country, language, pageview_count,
      events: [
        { event: "docs.pageview",     at, path: "guides/quick-start" },
        { event: "docs.page_feedback", at, path: "guides/quick-start", details: { vote: "down" } },
        { event: "docs.pageview",     at, path: "faq" },
        …
      ] }

visitor_id is a salted SHA-256 hash of the visitor's IP scoped to your workspace. It's stable across sessions (the same person visiting next week gets the same id), but raw IPs never leave Axiom and are never exposed via MCP.

Use cases:

"Who looked at our pricing page three times then dropped off?" → filter get_top_visitors, then get_visitor_activity for each candidate.
"This support ticket mentions a confusing page — let me see their full journey." → grab their visitor_id from get_page_journeys and drill in.
"Are AI questions and pageviews coming from the same people, or different cohorts?" → cross-reference get_visitor_activity across visitors.

Only events that carry server-side IP (docs.pageview, docs.page_feedback, landing.cta_click) can be attributed to a visitor. Pure client-side events (theme toggles, code copies) are visible in aggregate but not per-visitor.

See your events data live. Connect your GitHub repo →

Updated May 25, 2026