Crawl → referral
See which pages AI crawlers fetch, and which of those pages later send you humans. Cited-but-never-clicked becomes a list, not a guess.
How it works
The browser script counts people. The server middleware counts everything that never runs JavaScript. The dashboard — and your agents — read both.
A 1 KB tracker with no cookies, no localStorage identifiers, and no fingerprinting. It sends the page, the referrer host, UTM parameters, and a screen width. That is the whole payload.
<script defer data-domain="example.com"
src="https://analytics.example.com/js/v5.js"></script>Crawlers and agents never load your script. One middleware forwards their requests — user agent, path, referrer — to your Vector 5 instance. By default only non-human actors are recorded; set recordHumans for no-JS sites.
import { vector5 } from '@vector5/middleware/next'
export const middleware = vector5({
endpoint: 'https://analytics.example.com/api/hit',
domain: 'example.com',
})Open the dashboard, or point Claude, Cursor, or Codex at the MCP server. Six read-only tools: stats, pages, sources, AI traffic, goals, insights. Same aggregates the dashboard sees, nothing more.
{
"mcpServers": {
"vector5": {
"command": "npx",
"args": ["@vector5/mcp"],
"env": {
"V5_API_URL": "https://analytics.example.com",
"V5_API_KEY": "v5_live_…",
"V5_SITE_ID": "example.com"
}
}
}
}Everything else
Vector 5 is a complete replacement for your current privacy-first tool — and then it keeps going.
See which pages AI crawlers fetch, and which of those pages later send you humans. Cited-but-never-clicked becomes a list, not a guess.
People on your site in the last five minutes — humans and humans via AI only. Crawlers never inflate the live number.
Right now
27
Name an event, pass optional props, filter by actor. Conversion rates that agents cannot distort.
// Custom event with props
window.v5('signup', { plan: 'pro' })
// Ignore your own visits
localStorage.v5_ignore = '1'UTM parameters as first-class dimensions — utm_medium=ai_search is recognised as an assistant channel. Country and region from the IP, then the IP is gone.
An MCP server and a set of agent skills: audit crawler access, find cited-but-not-clicked pages, write the weekly founder briefing — from observed numbers only.
A live, public list of every AI crawler Vector 5 recognises and the source it maps to — GPTBot → ChatGPT, ClaudeBot → Claude, Google-Extended → Gemini. Useful even if you never install anything.
GET /api/crawler-index
Bring history in from CSV. Export every aggregate as JSON whenever you like. Your data is yours; the schema is documented in the repo.
POST /import · GET /export
Early access
Vector 5 is in private early access. Leave your email — we invite people in small batches. No account is created until then.
Self-host from GitHubInvite only · No public signup · No credit card