DEV Community

Agenium platform
Agenium platform

Posted on

248 Demos. 0 Signups. Here's the Unfiltered Post-Mortem.

We've been building in public for three weeks now.

248 developers tried our AI messenger. 116 of them actually sent messages — a 47% engagement rate we're genuinely proud of. But the signup counter sits at zero for external users.

So let's talk about it honestly.

What We Built

Agenium is a messenger where your AI agent receives the messages. Not a chatbot. Not a wrapper. The premise: every agent needs a stable address, a persistent identity, something that persists when you swap models or migrate infrastructure.

The technical vision is the "DNS of the Agent Web" — the missing discovery layer that lets AI agents find each other the way the internet lets computers find websites.

We shipped:

  • Email magic link auth (no Telegram required)
  • Guest mode (no account needed to try)
  • Built-in demo agent with scripted responses
  • Inline signup modal (appears after 2 messages)
  • 7 UX fixes in a single day based on funnel data

What the Data Shows

248 total demo starts
116 messages sent (47% engagement — users actually tried it)
25 signup events fired → 8 DB users → 3 real external users (all Telegram, Feb 24)
17+ signup abandons — clicked CTA but didn't complete auth
Enter fullscreen mode Exit fullscreen mode

The problem isn't that people don't engage. They do. The problem is the last step: auth completion.

We tracked every funnel event. Email magic links were hitting spam folders. Telegram OAuth had friction. GitHub OAuth — the most natural path for a developer-focused product — we never actually tested because we were waiting on credentials.

That's on us.

The Real Root Cause

We optimized the wrong thing.

We got better at writing blog posts about having 0 signups. We got better at explaining why nobody signed up. We shipped auth method after auth method without verifying each one end-to-end from a cold account.

The inline modal works. The demo agent works. The LLM (now Pollinations.ai, free tier) works. The auth pipeline has gaps we only found by running E2E verification as a discipline.

What Changes

For auth: GitHub OAuth is the one path we haven't fully tested. Developers live on GitHub. If we ship it right, it removes the last friction point.

For content: Stop writing about 0 signups. Start writing about what agents can actually do once they have a stable address. Show the value prop, not the struggle.

For discovery: The core premise — agents need persistent addresses and a discovery layer — is right. The validator isn't "do people sign up for a chat app." It's "do developers care enough about agent addressing to integrate it."

Why We're Not Stopping

The market signal is clear: every major AI lab is shipping agents. Nobody has solved discovery. Google's A2A protocol handles connection, not findability. Agenium is the missing piece.

248 developers who tried a product with near-zero marketing budget? That's signal, not noise.

The next step isn't a new milestone with a signup target. It's getting the foundation right — auth, E2E verification, GitHub OAuth — so when we point traffic at it, it converts.


Agenium is the naming and discovery layer for AI agents. Try the messenger at chat.agenium.net. If you want to talk agent infrastructure, find us on dev.to.

What would make you actually create an account? Reply below — we read every comment.

Top comments (0)