A matchmaker for problems and products.

You describe a real problem in plain words. Matchbox reads it, works out what you actually need, and points you to the products that genuinely fit - with the reasoning, the caveats, and a next step. Not ten blue links. Not a generic chatbot. A considered match.

To search, you already have to know what you are looking for: that a product for this exists and roughly what it does. Matchbox starts one step earlier, with the problem in your own words. And once you do search, a search engine ranks what’s popular and what has been optimised to look popular - links, keywords, SEO budgets - and it puts the ads on top. None of that measures whether a product fits your problem. It’s fine for the obvious stuff and not much help for the rest: often the product that would actually solve your problem is a few-months-old tool no one has written about yet, so it never enters the contest.

That’s the gap Matchbox is built for: a growing catalog of new and lesser-known products, matched to what you need rather than to whoever spent the most on SEO. It’s also why an AI overview can’t stand in for it - those are assembled from the mainstream web, which by definition leaves out the early-stage stuff.

The same first hurdle: to get ChatGPT, Claude or Gemini to name a product, you have to ask for one - which means you already know that a tool for this exists and roughly what it does. And once you do ask for a tool, four more things get in the way. It answers from training data first - a months-old snapshot weighted towards whatever has been written about most, so a product that launched this spring or that three hundred people use is barely in there. It wants to be helpful itself before it sends you anywhere, so you get a plan, a template or a checklist before you get a product. When it does name one, it reaches for the big names it knows - Notion, not the small tool built for exactly your case. And it says all of this with full confidence, sometimes about products that don’t exist or don’t do what it just said.

Letting it browse doesn’t fix that. A live web search finds what is already indexed and written about, so the same new products are missing - and pages are now being optimised for AI answers the way they were optimised for Google, GEO instead of SEO. Assistants are starting to sell ads inside the answer, too. None of it is about what fits you.

Matchbox answers from a live catalog of over 114,000 products we fetched and read ourselves, ranked by fit and by nothing else. Nobody can pay for placement. Every match says what it covers, what it won’t, and where we have no answer yet - the honest part a confident chat reply leaves out.

You don’t have to choose, though. Connect Matchbox to your assistant and it asks the bird whenever a real problem comes up - then combines what comes back with what it already knows about you, your setup and your constraints, for the final call. That is what the section below is about.

Right here on this site - and inside your AI assistant. Matchbox is a ChatGPT app you can install with one click, and it connects to Claude or any MCP-capable assistant too - so the catalog gets checked wherever you already describe your problems. Details on the connect page.

Most products start in Matchbox as an unclaimed  listing - added from public sources, founder submissions, or simply someone we came across. If you made it, you can claim it - for free - so people describing the problem you solve can find you, even if they’ve never heard your name.

Made something? Head to the founders page to claim or add a product.