How do I track the AI agents that visit my website?

The three observable sources

A referral can arrive with an assistant host as referrer. A fetch appears in an available origin log with a claimed user agent and, where supported, an IP verification result. A WebMCP tool call can be observed in the browser when the snippet instruments its API.

Existing analytics can measure identifiable referrals and custom events when configured. Compare the sources separately because they may overlap and have different coverage.

Set up browser signals

Sign in, add the domain, install the script and verify the installation. The snippet sets no cookie and does not retain a raw visitor address. It can record identifiable referrals and supported browser tool activity; crawler verification and remote tools require separate integrations.

<script defer data-domain="example.com" src="https://agenttracking.co/agent.js"></script>

Add the server log for crawlers

Upload an origin access log or send it through the authenticated log endpoint. Supported crawler identities are checked against current published ranges when available; unsupported or missing verification remains explicit. Close requests may be grouped into a burst, which is a pattern, not a known user question.

curl -sS -X POST https://agenttracking.co/api/logs/example.com \
  -H "Authorization: Bearer wmt_your_token" \
  -H "Content-Type: text/plain" --data-binary @/var/log/nginx/access.log

What you can see afterwards

Recognizable assistant referrals, log-observed crawler requests with verification state, and instrumented tool calls. Browser goal markers are attempts; site-server receipts are separately reported confirmed outcomes. API and MCP views preserve those evidence limits.

In short