AI agent detection.
Know who visits.
Tell people and AI agents apart through browser behavior. One tag, with no puzzles for your visitors.
/pricing → /signup
Software at the controls
- Pointer
- Direct jumps between targets
- Typing
- Even gaps between keys
- Scrolling
- Single jumps down the page
A browser can look human.
Its behavior tells you more.
An AI agent can use a familiar browser and user-agent string. Marshy looks at how the visit moves, types, clicks, and scrolls, alongside environment signals. Agent Detection-1 combines that evidence before calling a visit.
- Pointer and clicks
- Movement corrections, hover timing, and where clicks land provide evidence about what is driving the browser.
- Typing and scrolling
- Keystroke timing and scroll gestures help distinguish physical input from software-driven actions.
- Browser environment
- Automation flags, request headers, and signatures from agents that identify themselves add context.
From one tag to a useful verdict.
- 1
Connect your website
Create a project, add your domains, and copy your install snippet. The publishable key is scoped to your website.
- 2
Let the visit unfold
The tag collects behavioral signals in the background. Short or quiet visits can stay undecided.
- 3
Decide what happens next
Inspect the session in Marshy or listen for the verdict event in your app. Your own code controls the response.
Detection is the signal.
Your app makes the decision.
Agents can be useful customers, automation, or unwanted activity. A verdict identifies how the browser is being driven; it does not establish authorization or malicious intent.
Use it alongside your existing access and security controls. Browser-reported signals can be forged, so a verdict should inform your decision rather than replace those controls.
Start by understanding your traffic.
Label visits, explore the pages agents reach, and inspect session journeys before changing an experience. Marshy gives your team a view of both people and agents.
Explore AI agent analytics →What is AI agent detection?
AI agent detection classifies whether a person or software is operating a website. Marshy uses Agent Detection-1 by Resemble to evaluate browser behavior, report a visitor class, and make the verdict available in your console and on the page.
Does Marshy automatically block AI agents?
No. Detection is advisory. Your application decides what to do with a verdict, whether that is labeling a visit, offering an API, or adding a challenge on a sensitive action. A legitimate agent may be acting for a real customer.
What if a visit is too short to classify?
A visit may end before there is enough interaction for a verdict. Marshy keeps undecided visits visible instead of treating them as people or agents. Borderline verdicts can be marked for review.
Does the tag record what visitors type?
Typing signals record key categories and timing, not the actual text or passwords. Page content is not recorded, and visitor IP addresses are hashed before storage. See the developer documentation for the signal and retention details.
Meet the agents on your website.
Connect your site and see how people and agents browse. Sign in with your Resemble account to create your first project.