It tells you whether each session on your site is a person or an AI agent, and lets you decide what agents are allowed to do there: block them, charge them, let them through, or show them a challenge.
By how a session behaves, not by what it claims to be. Agents run in real browsers and can pass as people at a glance, but they still don’t act like people, and Shirimlabs picks up on that as the session happens. We don’t share the details, since that would make it easier to hide from.
Personal agents (like ChatGPT agent and Claude computer use), browsers with an agent built in (like Comet and ChatGPT Atlas), and the frameworks developers build agents with (like browser-use and Stagehand). The benchmark lists every agent we have run and caught.
Create a free account, add your site, and follow the setup steps in your console. Sessions start showing up there straight away.
No. People using your site never see Shirimlabs, unless you choose to show agents a challenge or a wall. Detection runs quietly in the background.
No. We never collect what people type into your site. See the privacy policy for exactly what we collect.
Yes. The playground is a test site with Shirimlabs switched on. Point any AI agent at it and watch it get flagged.
- Free. Block up to 500 agents a month on one website. No credit card. Start free.
- Pro. Block up to 10,000 agents a month across up to 3 websites, write custom rules, and charge AI agents for access.
- Enterprise. Limits set to your traffic, custom rules built with our team, API protection, security events streamed to tools like Splunk or Datadog, and direct support. Get in touch.