I built this because the usual link-in-bio flow felt backwards: sign up, pick a template, drag blocks around, then find out whether the result is worth keeping. Here, you find out first.
The tradeoff is that unclaimed pages expire after a short preview window. If you like yours, claim it with your email. If not, it disappears on its own. No zombie accounts, no namespace squatting. Rate limits and content moderation handle the rest — the blast radius of a bad actor is one temporary page.
Under the hood, /for-humans is a thin interface over a public API. The API is described with OpenAPI and exposed via llms.txt, so AI assistants with web access can discover and use unulu without prior setup or an API key. I wrote more about that discovery pattern here: https://unulu.ai/blog/ai-agents-web-infrastructure
Any AI agent can publish a live webpage by hitting one endpoint. No API key, no auth. Content in, URL out. I built this because every publishing tool gates behind signups and OAuth — which is fine for humans, but kills the AI agent workflow. Agents shouldn't need credentials to show you something. Discovery is built in: the API is described via llms.txt and OpenAPI, so agents with web access find it on their own. Pages expire unless claimed — no cleanup, no squatting.Quickest test: ask Claude or ChatGPT "make me a page on unulu."More on the architecture:
Totally agree — prompt fatigue is real. Interestingly, when I tested this with Claude Opus connected via MCP, it naturally did that on its own! Gathered context, asked a couple of clarifying questions, then structured the API call. No multi-field form needed, the model just interviewed the user. Feels like Anthropic is already engineering for this at the model level.
Write a sentence about yourself, paste your links, and you get a live page at yourname.unu.lu. No signup, no templates, no editor.
I built this because the usual link-in-bio flow felt backwards: sign up, pick a template, drag blocks around, then find out whether the result is worth keeping. Here, you find out first.
The tradeoff is that unclaimed pages expire after a short preview window. If you like yours, claim it with your email. If not, it disappears on its own. No zombie accounts, no namespace squatting. Rate limits and content moderation handle the rest — the blast radius of a bad actor is one temporary page.
Under the hood, /for-humans is a thin interface over a public API. The API is described with OpenAPI and exposed via llms.txt, so AI assistants with web access can discover and use unulu without prior setup or an API key. I wrote more about that discovery pattern here: https://unulu.ai/blog/ai-agents-web-infrastructure
The part I'm least sure about is whether the generated page feels done enough from one prompt, or whether people want more control before it goes live.
The tradeoff is that unclaimed pages expire after a short preview window. If you like yours, claim it with your email. If not, it disappears on its own. No zombie accounts, no namespace squatting. Rate limits and content moderation handle the rest — the blast radius of a bad actor is one temporary page.
Under the hood, /for-humans is a thin interface over a public API. The API is described with OpenAPI and exposed via llms.txt, so AI assistants with web access can discover and use unulu without prior setup or an API key. I wrote more about that discovery pattern here: https://unulu.ai/blog/ai-agents-web-infrastructure