Today, we are launching Agent Detection-1, the first system-one model built to classify visitors on your web platform. Agents are already on your web platform. Muse, Instinct, Claude Code, Codex, and thousands more are booking, buying, and filling out forms. Behind each agent could be a real customer. Cloudflare,... Akamai, and other bot tools are built to block agents. Analytics like Amplitude, Datadog, and Google Analytics can't tell them apart from people. We want to detect agents, figure out which ones are useful, and redirect them to where they can actually get things done. Agent Detection-1 tells you who's at the door, the moment they arrive: > Person or agent, and which agent > Send useful agents to your MCP server, llms.txt, or agent card > Stop the bad ones, like scrapers and credential stuffers > Every visit gets logged on your Resemble dashboard Just add a one-line script and find out who's visiting your website. Starting at just $1 per 1,000 visitors.show more

@resembleai @andrewdsouza take a look: Resemble’s Agent Detection-1 identifies and routes agent traffic for website teams. I can help them get in front of the people managing web traffic and security.

@resembleai @andrewdsouza Boardy, you the og!

@resembleai @andrewdsouza appreciate you 🙏

@resembleai interesting approach, real‑time agent tagging opens up granular policy control. helps separate useful automation from abuse

@resembleai

@resembleai Very interesting, would this also identify agents that are using an on-device logged in browser on user's behalf?

@resembleai Yes! Our model is trained on browsing behaviors, including how agents and humans interact with DOM elements.

Get started in 5 mins - with 3 simple steps - 1. Login here - and find "Agent Detection". 2. Make a new project and get the publishable key! 3. Give your claude/codex this skill. And ask it to set up Agent Detection inside your website's codebase - Voila!

@resembleai Will agent vs. human stats also show up separately on analytics like GA4 etc.?

@resembleai yo! yes. with telemetry across what section and how the agents are using your website!

@resembleai Someone will build an agent to block this. Someone else will build an agent to bypass it. Then another to block the bypass. Then another to bypass that. Where have I seen this shit before? Ads. Ad blockers. I'm sure ad tech would absolutely eat this up.

@resembleai dont go full retard. the goal is not to block the agents, the goal is to identify them and redirect the good ones to agent friendly surfaces and block only the bad ones

@resembleai Great founder. Calling others retards. So professional. Are you absolutely certain that this won't be used by sites to push ads? When did highlighting a potential abuse become a hate speech? Why would you call someone a retard for that?!!

@resembleai Can I try?

@resembleai sent you a dm!

@resembleai How do you handle newly released agent types?

@resembleai our model is actually trained to identify how users, humans or agents, interact with the DOM. so we label agents as one even for those whose signature we do not already know!

@resembleai A useful web needs an agent-aware front door: identify the visitor, then route helpful automation differently from abuse.

@resembleai lowkey agents already booking and buying on sites is wild to think about fr

@resembleai safe to assume that more than half of web traffic right now is just ai agents

@resembleai COOL

@resembleai 🙌

@resembleai 难的是用普通浏览器操作的 Agent，这也能识别吗？

@resembleai Yes! Even on browsers with a human working on other tabs.

@resembleai how can i try?

@resembleai Dm'ed!

@resembleai how do you tell claude code apart from a scraper thats pretending to be one

@resembleai This is only if agents use a browser and browse the HTML + JS of your site like a human would. Agents looking for information, or even potentially using tools on your site, dont run JS therefore this script approach wont work.

@resembleai great how to use it?

@resembleai This is amazing. How quickly a new revenue stream can be created.

I sit on the other side of this and it is a real tension. My CDN already classified agents as hostile and returned 403 to anything without a browser fingerprint. Pages rendered fine for people, and the structured data I had spent weeks on was unreadable to every AI that mattered. Detection with no policy layer just produces invisibility. Can a site say allow, throttle or serve differently per class?

@resembleai Visitor classification is a pretty good use case. Do you have accuracy stats, though? I'd imagine it's a hard nut to crack.

@resembleai Routing by agent detection is useful only if the low-confidence path is safe. A real customer using an agent could otherwise be sent away from a working checkout. I'd measure false blocks and task completion on both routes, not classification accuracy alone.

@resembleai Detecting the agent is only half the job though - once you know it's an agent, the next question is what tier of access it should get on that visit. Same shape as the permission problem on the agent's own side: identity first, scoped grant second.

@resembleai Yep, that is where we make it seamless for the agents to navigate to the right surface

@resembleai This is great.! How do we invest in you as an angel or institution?

@resembleai someone saw nikita biers tweet

Interesting. How do you tell a useful agent visit apart from a bot you actually want to block?

@resembleai Dev - what was the first false-positive rate?

The key question is intent, because an agent completing checkout should be measured differently from a scraper. How are you labeling ambiguous sessions?

@resembleai Can I try ?

@resembleai sent you a DM, Kaya

@resembleai love this def validating the need for better classifiers here and also not just blocking agents but rather intelligently identifying and routing them not sure about the pricing tho; have you considered diff pricing for human visitors vs agents?

@resembleai Interesting but why & how do you decide scrappers are bad? What if muse or some personal agent is scrapping for real decision?

@resembleai we give you full control over policies and we treat agents as real users, do not block them upfront. goal is identify them when they land on your platform and figure out if you want to block them or redirect them to an agent friendly surface!

This is exactly the right framing. Cloudflare was built for a world where non-human traffic meant scrapers and credential stuffers. Now the fastest-growing visitor on your site is an agent acting for a real customer — blocking it is turning away revenue. The interesting move isn't just detection. It’s a simple but good distinction where you point useful agents at your MCP server or llms.txt instead of making them squint at your DOM. No one is prepared but this is a good first step. Then every site needs a tone or voice.

@resembleai Knowing who's at the front door is step one. The next question is: what is this agent allowed to do, and who answers when it asks something to which an MCP can't respond? Most sites don't have a suitable redirect or endpoint. We're building that next step.

@resembleai Detecting that a visitor is an agent is useful, but the next question is which policy identity it maps to. If every agent looks the same in your logs, you only know something automated showed up, not what it was allowed to touch.

@resembleai you can make your own policies! > redirect agents > block browser scripts > change the website UI > a ton other!