AI agent analytics.
See the whole mix.
Measure agent traffic, explore session journeys, and understand who is behind the visits on your website.
AI agent share of all sessions
25%
A visit isn't always a person.
An assistant comparing plans and a person reading your pricing page can both create a session. Marshy puts the visitor classification beside your traffic metrics, so your team can study human activity and agent activity separately.
Keep undecided visits in the picture.
Agent share is calculated against all sessions. Visits without enough evidence stay visible, rather than being silently counted as people.
Example: 250 agent sessions / 1,000 total sessions = 25% agent share
Follow the pattern over time.
Compare daily totals and agent share. A rise in sessions means something different when most of the additional visits are software.
Start with the traffic mix, then open the sessions behind a change.
Go from totals to actual visits.
Find the pages agents use
Explore page activity and sources to see where agents arrive and what they browse. A referrer is context, not proof that a visit is an agent.
Inspect the session journey
Open a session to see its page journey, visitor class, and the evidence behind its verdict. Follow the individual visit instead of guessing from a traffic spike.
Connect visits to your users
Identify signed-in users with a signed ID. Review each user's agent share and open the sessions associated with their activity.
Understand the kind of automation
Explore visitor classes such as computer-use agents, browser automation, and scripted clients. Different kinds of software can need different experiences.
Classification before interpretation.
AI traffic analytics starts with understanding how a visit is driven. Marshy uses browser behavior to distinguish people and agents. It does not infer the visitor's intent or authorization from a referral link, and it does not automatically block anyone.
See how AI agent detection works →What does AI agent analytics measure?
Marshy measures how much website traffic is people, AI agents, or undecided visits. The console lets you explore daily trends, pages, sources, agent types, sessions, and identified users.
How is agent share calculated?
Agent share uses all sessions as the denominator, including undecided visits. For example, 250 agent sessions out of 1,000 total sessions is a 25% agent share, even if some of the other sessions have no verdict.
Can I see which signed-in users have agent activity?
Yes. Identify users after sign-in and sign their IDs with your server-side identity secret. Marshy can then connect sessions to users and show their share of agent activity. Keep the identity secret out of browser code.
Can I send verdicts to my existing analytics tool?
Your application can listen for the resemble:verdict event and send classification data through its own analytics code. Native third-party integrations are coming soon; the current integration catalog describes planned connections.
Put agent traffic in context.
Add the tag to your website and explore the people, agents, and undecided visits in your console.