Integrating Bubble and Claude

Questions
1

Hi all, I’m working on an app where I’m using the Claude API to send it data and give it some instructions and then retrieve Claude’s response and display it to my users.

So far I’ve just been sending data that my users have just entered in the app. But I really need to take it up a level now and allow Claude to have access to data that my users have entered previously, and, indeed, to give Claude access to some of its own previous responses.

Does anyone have any ideas on how best to do this?

My current plan is to use the new Files API feature in Claude in order to send it text files containing the historical data. I think that should work fine but it also feels a little cumbersome and I’m going to need a lot of logic to figure out which files I need to send it and tell it to use.

So I’m also looking at this MCP concept, which, as I understand it, could be used to give Claude access to the Bubble database, I think.

Is there such a thing as a Bubble MCP server? Or any plans for one? Or is that something I could build myself? Does anyone have experience with this?

Another alternative I’m considering is storing my Bubble data in Xano and using their MCP server to give Claude access. I just watched a Xano video and it looks possible. But obviously then I’m giving myself another task in deciding which data I need to store in Xano etc.

Any thought, recommendations on this topic gratefully received!
Thanks
Tony

2

I doubt Bubble will be building any MCPs for DB access. Last thing every Bubbler wants is AI destroying the shared cluster.

I advice using an external DB for this. You will thank yourself in the future.

You don’t even have to migrate your DB from Bubble (and make sweeping changes to your Bubble app). Since you only need specific data to be accessible, just duplicate data to Xano. A little extra steps but now you have a modular setup and makes troubleshooting between the Bubble app and your AI workflow a lot easier. Just remember to keep it synced if necessary.

You can even schema it to be more in line with your AI workflow. An optimized AI workflow will save you a lot of money in the long run.

3

This tutorial shows you how to do what you are describing.

You don’t need to MCP or any external database.

4

It depends. Can you do it with just Bubble and some API calls and some workflows? Yes.

The problem comes when you need scale or/and frequency. Imagine paying for WU for every read from Bubble’s DB to 1 prompt.

For OPs case we’re looking at using Bubble’s DB as memory for AI. So how do you picture this playing out?

5

I must be missing something.

I read this as simply needing to add more context to the prompt.

Optimal setup for scale is a different question but not sure it’s part of the equation :slight_smile:

6

Thanks @ihsanzainal84 and @MattN for your replies, very helpful.

Apologies if I didn’t explain the use case clearly enough but I guess what I really need is a kind of knowledge base feature for my users. The knowledge base will initially be made up of their previous interactions with Claude. But in future it could include text uploaded directly by the user, eg articles or documents that they’ve written.

Sounds like setting that up in Xano is the way to go. If you have any other suggestions or advice for things I should consider I’d love to hear it. :slight_smile:

Thanks!

1 Like
7

You might want to consider this to allow your LLM to answer with data from your database

8

Interesting, thanks. I’ll take a look.