Wholetally
Веб-аналитика без файлов cookie для агентств, доступная для запросов из вашего ИИ-ассистента
Размещённый MCP-сервер
npx add-mcp 'https://api.wholetally.com/mcp'Устанавливается в Claude Code, Codex, Cursor и другие
Документация
See which AI reads your sites. And the visitors GA4 misses.
Wholetally shows which AI crawlers read which pages and which visitors ChatGPT, Perplexity or Claude send you. It also counts the people who decline cookies, block trackers or run no JavaScript, who never reach a consent-gated GA4. No cookies.
No card needed. The demo is a sample dashboard with example data, no sign-up.
Running several client sites? One Business account covers 30 of them. See what agencies get.
Where GA4’s count stops.
Four kinds of traffic a consent-gated, script-only setup can’t see. Three are the yellow bands in the chart above. AI crawlers get their own count, outside your visitors.
GA4 Wholetally
Said no to cookies GA4: Basic consent mode sends nothing. Gaps are modelled only above traffic thresholds. 1 2 Wholetally: Counted. No cookies, nothing stored on the device.
Ad or tracker blocker GA4: The Google tag is on EasyPrivacy, on by default in uBlock Origin. 3 4 Wholetally: Counted. Script on your own domain, edge as backup.
No JavaScript ran GA4: Needs its tag running in the browser. 5 Wholetally: Counted at the edge, from the HTML response.
AI crawlers GA4: Filtered out. You can’t see how much. 6 Wholetally: Counted per page, kept out of visitors.
Sources (6), checked Nov 2026
Count the page load, not only the script.
A cookieless script under 2 KB runs from your own domain, so fewer blockers catch it. Add the edge collector, a one-file Cloudflare Worker or Next.js middleware, and every HTML page load counts, with or without JavaScript.
When the script fires too, its hit wins and the edge hit is dropped. The docs say where matching can miss.
Visitor Any browser, blocker or not
Your domain Cloudflare Worker or Vercel middleware
HTML from your origin 1 beacon after the response
Wholetally collector EU, deduplicated against script hits
Keep AI crawlers apart. See who AI sends you.
GPTBot, ClaudeBot and the rest read your pages without running a script. They are counted per page at the edge, outside your visitor numbers. Visitors from ChatGPT, Perplexity or Claude get their own channel.
example-shop.nl Example data Last 30 days
AI crawlers 8,494 requests, counted at the edge
| Crawler | Purpose | Most requested | Requests |
|---|---|---|---|
| GPTBot OpenAI | Training | /blog/ideal-vs-creditcard/ | 3,184 |
| ClaudeBot Anthropic | Training | /products/ | 2,207 |
| OAI-SearchBot OpenAI | Search | /pricing/ | 1,342 |
| PerplexityBot Perplexity | Search | /blog/ | 961 |
| ChatGPT-User OpenAI | User request | /returns/ | 412 |
- chatgpt.com AI assistant 1,218
- perplexity.ai AI assistant 344
- claude.ai AI assistant 187
Follow a visit to the sale, when you switch it on.
Funnels show each step, the drop-off and which sources finish. They need the daily visitor hash, a per-site setting that is off by default: a salted hash that rotates every 24 hours, with only totals kept.
Connect Stripe and each payment is matched to the source of its visit. Your checkout passes the source along; no visitor ID.
How funnels work Revenue per source
example-shop.nl Example data Last 30 days
Blog to checkout 7.4%
- 01 3,120 Read the blog entered
- 02 1,486 View a product 47.6% continued
- 03 512 Add to cart 34.5% continued
- 04 231 Start checkout 45.1% continued
Last 30 days Previous 30 days
- Google 6.9% finish
- chatgpt.com AI assistant 13.1% finish
- perplexity.ai AI assistant 10.4% finish
Get a plain-language read of every chart.
Explain sits on each chart and writes a short reading from its aggregated numbers: what the metric is, what moved and what to look at first. It is AI-generated and uses only counts, never visitor-level data. Handy when you have to tell a client what a chart means.
example-shop.nl Example data Last 30 days
AI crawler requests Explain
- GPTBot 3,184
- ClaudeBot 2,207
- OAI-SearchBot 1,342
Explain AI-generated
The AI report keeps two things apart. Crawlers are bots that fetch pages: 8,494 hits in this window. GPTBot fetched the most, 3,184 hits. Separately, 1,749 visitors clicked through from an AI assistant, most of them from chatgpt.com (1,218). The visitors are already part of the 15,189 total, the crawler hits are not.
Example answer, written from the example numbers above.
Ask your own AI instead of opening a dashboard.
The built-in MCP server gives your assistant your visitors, pages, sources, AI traffic, funnels and revenue. Connect Claude or ChatGPT by URL and sign in with OAuth. A wt_… token still works for Cursor, Claude Code and VS Code.
Read-only, apart from adding a site. This is not a directory listing.
Terminal
# a read-only MCP token from Account → MCP tokens
export WHOLETALLY_TOKEN=wt_…
$ claude mcp add --transport http \
--scope user wholetally \
'https://api.wholetally.com/mcp' \
--header 'Authorization: Bearer ${WHOLETALLY_TOKEN}'
Which sources brought revenue last week, and did any come from AI assistants?
revenue_by_source { "site": "example-shop.nl", "period": "7d" } read-only
Google brought the most: €1,042 from 14 orders. chatgpt.com came second with €486 from 5 orders, the only AI assistant with sales this week.Example answer. Every number comes from the tool result.
This is the whole hit.
By default only totals: no cookies, no localStorage, no fingerprinting, no visitor ID, no stored IP. A visit is unique when it arrives from another site or with no referrer, the rule Simple Analytics uses (Simple Analytics docs: Unique visits).
Daily hash, only if you turn it on
Off by default. Needed for funnels. Not available in strict mode.
Open-source script
MIT-licensed. Read every line it runs.
Compliance modes
Per site: standard, strict, and Dutch and French modes that mark a site as configured for the local exemption. Legal review is pending, so check with your own legal advisor.
POST /_wt/api/event // your domain
{
"t": "pv",
"s": "example-shop.nl",
"u": "https://example-shop.nl/shirts/linen",
"r": "https://chatgpt.com/",
"w": 390,
"l": "nl-NL"
}
t, s, u
Pageview, your site, the page
r
Where the visit came from
w, l
Screen width and browser language
No cookie, no storage on the device, no visitor ID. The IP address is not stored.
A new product, with its workings in the open.
Built in the Netherlands
Data is stored in the EU. Company details and the processing agreement are on the legal pages.
Claims with sources
Every statement about another product on the compare page links to the page that says it.
Public changelog
What shipped and when, in the changelog.
Measured with itself
This site runs the same cookieless script. Every field it sends is listed.
Priced by the pageviews you count.
Starter from €7 /month 1 website · 1 member One person, one site. Everything you need to read your traffic.Growth from €12 /month 5 websites · 5 members A small team. Funnels, the edge collector and the MCP server.Business from €19 /month 30 websites · 25 members Agencies and product teams. Revenue per source and the full API.
Starting prices at 10k pageviews a month, in euros excl. VAT. Yearly billing gets two months free. Every plan starts with a 30-day trial, no card needed.
Count the visitors you’ve been missing.
30-day free trial, no card needed. Add your first site in a few minutes, or look around the demo first.
Footnotes
-
Google Tag Platform: Consent mode overview when the user doesn’t consent, no data is transferred to Google at all -- not even the consent status. ↩
-
Analytics Help: Behavioral modeling for consent mode at least 1,000 events per day with analytics_storage=’denied’ for at least 7 days. ↩
-
uBlock Origin: default filter list configuration "easyprivacy" … "group": "privacy" (no "off": true, so enabled by default) ↩
-
EasyPrivacy filter list (tracking servers) ||googletagmanager.com^ ↩
-
Measurement Protocol (Google Analytics 4) The intent of the Measurement Protocol is to augment automatic collection through gtag, Tag Manager, and Google Analytics for Firebase, not to replace it. ↩
-
Analytics Help: Known bot-traffic exclusion you cannot disable known bot traffic exclusion or see how much known bot traffic was excluded. ↩