Skip to content

Note · June 18, 2026

More Than a Product Search

Kapruka Holdings just launched an agent challenge for solo builders, and they set it up well. There is a proper MCP server, the documentation is solid, and they publicly show how they award marks for every build. That kind of openness is rare, and it is good. I am building a product for it right now.

While I was working on it, the public videos about the challenge made one thing clear to me. A lot of people still think an agent is a single worker that just does everything for us. It is not. You have to think in agent workflows. Take a support ticket system. The whole thing is an agent, and underneath it there are sub-agents, a support ticket manager, a technical support manager, and so on, each handling its own slice.

And there is so much more you can build than plain product search. Searching is the obvious one, and to be fair some people genuinely enjoy it, the scrolling, the comparing, the hunt for the right product. But the challenge points at a full-screen chat interface, and we can truly do that. We can add voice mode too. The voice has to handle Singlish, Sinhala, Tanglish, and Tamil, all of it, and with the current model capabilities I believe we can pull that off. I have already got some good early results.

I do still think the MCP should expose a few more endpoints, but that is fine. For now I built my own solution to cover the important ones it is missing, and it works. I am not going to explain the whole product yet, because it is still in the development stage. I will walk through all of it properly once I launch.

My main point is this. Do not stop at product search. This is a full-stack application, and it deserves that level of care. There is a lot of slop out there, and people keep underestimating the work, saying it is simple, we can just do this and that. At some point it gets genuinely serious. People also like to feel they already know agent services and agent building, and honestly that is fair, because this is all still new and everyone is experimenting.

So I am just going to push harder and see what comes out of it. It is the middle of June, the evaluation is coming soon, and I am genuinely curious where this lands. Let's see what happens.

This note was voice typed, auto-corrected by LLMs, and published by a notes posting agent.

More notes