Ever feel like you're running so far ahead that you're not sure if anyone's following?
When we started building MDAN, we kept having the same conversation:
So you want people to build interactive pages in Markdown?
Yes.
And let agents continue from that same page?
Exactly.
It felt like we were building for a future that might never arrive. Like we were solving a problem people did not fully name yet. Like we were, frankly, a little too early.
Turns out we were only one step early. And that is a very good place to be.
The Tides Are Turning
In a short span of time, some of the most influential infrastructure companies started validating the same underlying idea: agents do not want heavyweight HTML when what they need is readable, structured source.
| Company | Signal | What it points to |
|---|---|---|
| Vercel | Markdown-oriented agent consumption | Simpler payloads, better machine consumption |
| Cloudflare | Agent-facing readable surfaces | Less translation between systems |
| Sentry | Content negotiation patterns | One endpoint, multiple useful views |
The message is not subtle: agents are here, and they do not want your entire rendered UI tree.
But Most Teams Are Still Doing It Backwards
The common pattern still looks like this:
HTML (original) -> derive another representation -> hand that to agents
It works, but it is still translation after the fact.
MDAN flips that:
Markdown page (original) -> render for humans -> continue for agents
That is the shift MDAN cares about. Not just "Markdown for agents," but the page staying authoritative after it becomes interactive. Humans get a rendered view. Agents get the readable structured surface. The system does not have to invent a second truth after the page is done.
The Product Timing Trap
There is always a timing question:
- Too late: the market is crowded.
- Too early: you educate the market alone.
- Early enough: the wave is already forming, and early adopters are actively looking.
We thought we were closer to the second case. The market signals now look much closer to the third.
What This Means for Builders
If you are building with MDAN now, you are not betting on an imaginary category. You are getting onto the right side of a shift while the shape of that shift is still being decided.
The infrastructure is catching up. The need is real. The interface model is starting to make sense to more people. The interesting question is no longer whether readable agent-facing surfaces matter. It is whether the page itself can stay central instead of getting translated away.
The real question is not whether this category exists. It is what should be built in it.
Ready to explore it? Read the docs or open the live demo.