Using my Custom GPT in a Website and Mobile Apps

I’ve developed a custom GPT here in ChatGPT, and I’m now looking to turn it into a standalone app. I’d like to know if anyone has experience integrating a custom GPT into a website or a mobile app.

Is it possible to embed or call a custom GPT in an external application? If so, what would be the best approach? Are there any recommended frameworks, APIs, or best practices for doing this efficiently?

Additionally, I’d appreciate any insights on making it accessible to users and potential monetization strategies.

Looking forward to any guidance or experiences you can share. Thanks!

2 Likes
2 Likes

Thank you for the clarification!

1 Like

Offering a nuanced approach if you are looking for a more involved application.

There is nothing preventing you from building your external application to allow your custom GPT to interface with it You could allow them to edit a conversation history stored through an API or modify settings. Think like picking up a written document from within the GPT canvas to being able to edit on your website.

The application would be like if you want them to use the custom GPT to interface with the settings/orders/general content.

1 Like

https://community.openai.com/t/using-my-custom-gpt-in-a-website-and-mobile-apps/1136500

Hi!

I saw your reply to Seco’s post and I am trying to do exactly the same. I’m surprised it didn’t get more attention!

Please could you kindly point me to any links or threads where there is more information / a tutorial on how to connect to (or replicate) a custom GPT’s functionality from a standalone web app?

For context: I am trying to transcribe lyrics if a user provides a Soundcloud link

Thanks very much!

I was just saying if you have a website similar to how you have a frontend then treat the customGPT as a front end.

The user logs into your website via the customGPT they upload their music in customgpt, or get file through an action, once it is transcribed and then you use a custom action to upload that to the web services. They then interact with the transcribed lyrics on your website afterwards. This could be similar to how you might have some audio plugins in protools open in separate window off to side in your workflow except it’s the chatgpt window not sure you might be able to load it in an iframe to have more seemless experience line its built into the website.

If they need to make changes they could update the document inline on your site and then go back to the GPT after making changes to several documents or just the one - go back to customerGPT and ask it to pull updates. It would query your service and pull the users changes and then proceed to make those changes and upload the document back to your website.

You can’t make the CustomCPT a standalone app but you can make it a companion app.

The key is just have a web service you can make custom actions for .

Hope this helps, wasn’t meant to mislead just look at the problem from a different perspective .

Thank you for taking the time to expand on the idea.

I’m still trying to digest your suggestion but it sounds like an iframe on a website showing my instance of the custom GPT in my OpenAI plus account.

I’m not an expert but would this workflow scale ok on a commercial website?

It sounds like it would work fine for one person but if you have hundreds of people somehow accessing my private custom GPT is there not a possibility that I could get rate limited and the service could be temporarily paused? Even on my own, I’ve had a temporary warning that the service has paused because I am using it too much.

What are the implications? I don’t mind paying API fees for the service if I can make a small charge to visit us for the service.

It would be really useful if there was some sort of tutorial for this type of business model. A lot of people are thinking about the possibilities to monetize their ideas.

Anyway, apologies if I’ve totally misunderstood. Thank you for your input!

i added my gpt instructions to github, multipersonagpt, maybe I can update there.

In short, you don’t login to your gpt your users login to your your website through your custom with oauth.

Lookup openai cookbook, custom gpt, middleware in a web search. Should be some integrations with AWS, Azure and Google cloud. Some people use make.com or zapier as well.

I can post some more details on the github unless you have a reference to a different thread you start here in the community.

I don’t want to hijack this thread any more than might already have.

1 Like

Thanks again. Going to have a look at this again when I’m fresher tomorrow.

By the way, I don’t think you’re hijacking the thread as it seems you are answering a similar question to what the original poster asked and it might help others who are seeking to do the same

Oh, well then. If you have a database and web api in front of it that used for you website.

The user logs into your website via oauth through ChatGpt CustomGPT. They have their own session. Whatever they upload to your web service could be “owned/related” by that person.

Depending on the roles and security you have set up for example if lyrics are viewed by everyone. Then in that a table whenever they make updates to it - it would be viewable through all users.

There is no scaling issues because everyone has their own independent session with your website and each customGPT is just a session. No different than as if they visited your home page and logged in.

You would need to decide on your access roles and work out when and how things were updated.

If I upload something from my chatgpt custom gpt session and then if another user uploads either exact same thing or slightly different which gets shown to all users. If you didn’t want to use a database you could use a object store and store things by checksums for filenames like <user uuid>-<song-id>.txt.

In your home page when all users are viewing all the songs maybe the see all entries in the bucket or maybe they only see entries related to their <user uuid>.

Similar back in the custom gpt your web service could allow them to pull any song from that bucket or just songs with their <user uuid>. It really comes down to how you intend to setup roles and permissions.

Everyones using the customGPT is going to come down to their plan so if they are on a plus plan will have those limitations. Usually, or many times, when you are designing a REST API you will design that Rest API with openapi/swagger defintion that the frontend gets tested against. There is no reason you can’t just copy and paste that openapi definition into a custom action for yoru customgpt so it can make all the same calls that your frontend would make.

There are so many way of doing everything it’s not to go down with a rabbithole.

Your charge would have to be your website fee to access web interface or or enable their login through the customGPT.

Rate limits don’t apply to your customGPT, in a a universal sense, that I am are of. It’s a benefit to ChatGPT and is adding value to the ChatGPT Plus by you offering it through ChatGPT.

To start with you could look at using API Gateway, Lambda, and S3 to put together something simple just to get a customGPT up and running and then add the frontend later.

Shameless plug, go into my MultiPersonaGPT and say “add a team of AWS expert developers to guide me through settting up a webservice with API Gateway, Lambda, and S3 that can be used with a ChatGPT CustomAction”.

Maybe I’ll add the IaC and you can just download it and run if you have AWS if you’d be interested. Feel free to ask specific quesitons I only mentioend didn’t want to hijack it out of respect for the orgiinal post incase they were getting notifications.

I think we need some type of reliable hosting, where OPENAI supports “US” CUSTOMERS, to serve CUSTOMERS. You know there is a word for such collaborations, named “PARTNERSHIPS” guess you forgot the meaning of those?