Manufact's Launch Reveals the Growth Play: Chat Apps Are Starting to Recommend Products to Users Based on Intent. Ship an MCP Before That Channel Fills Up.

Manufact's Launch HN post reads like a product pitch.

Underneath, it's a distribution forecast — and a fairly specific one.

Buried in the post is this: "Clients (Claude, ChatGPT, Cursor) are starting to dynamically present MCP servers/apps to users, based on their intent. Products will be organically discovered on the chats!"

Why this matters beyond Manufact

This isn't just one startup's roadmap. It's a claim about where a whole discovery channel is heading, backed by a number: "all major tech companies have an MCP server, and many of those are reporting that already 15+% of their usage comes from their MCP."

15+% of usage, today, for the companies that already built one. Most companies have not built one at all.

What the launch post got right

1. It named the real number

Most "AI distribution" advice is vague. This post gave a specific, attributed figure: an Amplitude engineer's MCP saw "a 2x increase in retention after adding UI to their MCP." That's not a prediction — it's a reported result, from a real product, tied to a specific change (adding UI).

2. It described the mechanism, not just the trend

Dynamic, intent-based presentation inside ChatGPT and Claude means a user doesn't have to know your product exists to be shown it. That's structurally different from SEO (you have to be searched for) or ads (you have to be paid for). It's closer to being recommended by the client itself, based on what the user is trying to do.

3. It flagged how early this still is

The post's own timeline shows how recent this is: "Oct 2025. ChatGPT Apps SDK," "Late 2025. The stores open," "Jan 2026. MCP Apps becomes official." That's a channel that's had store directories for roughly two quarters. Most companies haven't caught up yet — which is exactly the window worth moving in.

How to Build This Into Your Own Launches

• Ship an MCP server for your product now if you don't have one — the post states "all major tech companies have an MCP server" as the current baseline, and that bar keeps rising
• Design your MCP's tools around what a user is actually trying to do inside a chat, not as a 1:1 mirror of your REST API — the post specifically warns that kind of proxy MCP tends to get abandoned
• If your product has anything worth visualizing — data, results, a dashboard — build the interactive UI layer; the cited retention lift (2x) is directly tied to that one change
• Get your submission right the first time; the review queues for ChatGPT and Claude directories are still new and manual, and a rejection costs you time in a channel where being early is the whole advantage
• Watch your own MCP usage numbers the way the post frames it — "15+% of usage" is the benchmark companies with a working MCP are already hitting; track whether yours is trending toward that or nowhere close
• Don't treat your MCP as a one-time build — the spec is still moving (MCP Apps only became official in Jan 2026), so budget for maintenance the same way you would for SEO or app store optimization

What to Watch

Watch how fast the "15+%" usage figure moves for companies with MCPs already live, and how many net-new companies ship one over the next two quarters. If chat-based, intent-driven discovery keeps growing the way this post suggests, the founders who shipped a well-designed MCP early will own a distribution channel that latecomers will have to fight much harder to break into.

Source: https://news.ycombinator.com/item?id=48762862