I checked how Apps work and found that they function within ChatGPT, but it’s currently not possible to use Apps inside Custom GPTs.
Integrating Apps into Custom GPTs would be an ideal combination because:
Custom GPTs are great for text conversation.
Apps have an graphic interface that can be displayed in a separate message.
It would be really cool to combine them so that Custom GPTs have a richer interface. @OpenAI_Support — will it be possible in the future to combine Custom GPTs with Apps built using the Apps SDK?
I’m not OAI staff, but I’m pretty sure this Apps SDK is their way of sunsetting custom GPTs, not supplementing them. It seems to combine what worked with plugins and custom GPTs into one SDK. Which I, for one, fully support and am excited about.
I hope they won’t sunset custom GPTs — that would be really disappointing. Instead, they should allow the use of Apps, and even Apps within custom GPTs. That would create a much more flexible user experience and enable interesting combinations.
Also, if each type of product (Plugins, GPTs, Apps) is discontinued every two years, what’s the point of creating and growing them? It’s much easier to build outside the ChatGPT platform; otherwise, it feels like a waste of time. Products need time to grow and eventually monetize. But once they find product–market fit and achieve sustainable monetization, they get sunset and the cycle restarts — that’s not cool. It’s much easier to build outside their platform in that case.
Moreover, many people have custom GPTs tailored to their own specific needs - it’s not very convenient to rebuild everything from scratch each time a new release comes out.
I really hope they don’t sunset custom GPTs. But if they do, I’ll probably build my next products outside the ChatGPT platform, because in that case, developing and growing products here would seem increasingly pointless.
called it (that custom gpts wouldn’t work, some people got emotional on this but it was objective not subjective)
and to be honest, I’m not a fan of MCP, it’ll probably introduce a bunch of zero days for a bunch of companies. In my opinion, just do what was done with sora and codex and have it on another domain and call it the app store and full on host apps there
its such a difficult problem to solve too (in terms of product market fit, not an easy feature/project/repo to role out, so major props to those working on it from a business decision and dev perspective), but the apps approach at least seems way better then the custom gpts, just feels like a rebranding of connections imho
on topic response: just add a mcp to the custom gpt and make sure it follows the guideliness, maybe set up a new repo or rebase the current repo, shouldn’t take an absurd amount to get it there (and you have until the end of the year to do so)
Meanwhile, google is in the back saying ”Oh, you killed it after 2 years? Hold my beer.”
Correct . Many of us would agree with that sentiment. At the end of the day, all you need is a chat completions API. This tends to be the evolution of a developer. Plus, the more you architect your own infrastructure, the more resilient you become to such things, preventing you from needing to “rebuild” during each change.
very much agree with this, expecting to cover server costs with a customgpt or an app alone seems to only be possible once you grow and find product market fit on our your own. Once you have a pretty large base of paying users/companies is when you’d offer this.. starting at a customgpt or an app seems like a moon shot to me
https://www.youtube.com/watch?v=hS1YqcewH0c
. Many of us would agree with that sentiment. At the end of the day, all you need is a chat completions API. This tends to be the evolution of a developer. Plus, the more you architect your own infrastructure, the more resilient you become to such things, preventing you from needing to “rebuild” during each change.