Is there a tool that tracks only the agentic use of my website?
Agent Tracking groups identifiable assistant referrals, origin-log crawler requests and instrumented browser or server tool calls. These are separate types of evidence, not a count of unique autonomous agents. Browser goal markers are attempts; confirmed outcomes require a site-server receipt.
By Gorden Wübbe · Updated 2026-09-25
Why use additional data sources
Web analytics can segment identifiable assistant referrals and can receive custom tool events. Crawlers usually require server or edge logs. Agent Tracking brings these sources into one view while keeping their counts and evidence levels distinct.
What the product records
Recognized assistant referrers, log requests claiming supported crawler identities with explicit verification status, and instrumented tool calls. Google-Extended and Applebot-Extended are robots policy tokens, not crawler identities. Goal markers are unverified attempts. Plain page views provide context.
What it deliberately leaves out
No people analytics: no bounce rate, no funnels for humans, no heatmaps. No prompts, spans or token costs of agents you build yourself; that is LLM observability, a different product. No blocking: it measures and never interferes. The workspace combines measurement views, evidence-based insights, prepared browser tests and protected reports.
Open source, or hosted
The software is AGPL-3.0 and runs on your own server with one compose file, or as a hosted service in Germany with a free pilot. Either way the data is one SQLite file and the code is public.
In short
Does it count human visitors at all?
Only as a daily total beside the agent numbers, so you can see the proportion. Human visits are not classified, not segmented and not charged.
Can I get the numbers out?
Yes: JSON through the stats API, CSV export, a weekly mail, and an MCP tool your own assistant can call.