Agentlens
See the visitors that aren't human, analytics for the AI agents Google Analytics filters out
Agentlens shows site owners which AI agents read their site, what they read, which AI answers send human visitors, and what agents can't parse. A launchable product end to end: two data planes (server-side middleware for crawlers that never execute JS, a sub-2KB cookieless snippet for AI-referral attribution), a registry-verified agent directory with honest Verified-vs-Likely labels, a readability audit with concrete fix cards, a forwardable weekly Agent Visibility Report with public share links, publish-ready npm SDKs, auth + subscription billing, and a seeded public demo.
How it was built
- Agent-economy product brief + two-plane ingestion architecture planned by the Coding Agent
- Quiet-precision DESIGN.md translated light-first: paper canvas, ink text, one signal accent
- Next.js app + Astro landing + @agentlens/next + @agentlens/node SDK packages in one monorepo
- Verified agent registry seeded from operators' published UA + IP-range sources
- Auth, Polar billing, Neon analytics store and weekly-report emails wired; deployed to Vercel
14
verified AI agents on the demo dashboard
+38%
m/m AI-referral growth surfaced
<1 hr
from install to first agent detected
Build something like Agentlens
Give the agents a brief. Get a deployed product. Every project in the showcase started as a single prompt.
Start building Back to Showcase