Build an MCP of your Bubble App

[design.agx]

Hey there
I built a tool that lets you create an MCP server for your Bubble app.

Sebastian's video 🚨 — Tella

Once your MCP is set up and you’ve added the endpoints you need, you get a server URL that you can plug directly into Claude or ChatGPT as a custom connector.

From there, you can query your Bubble app in natural language, directly from Claude or GPT.

Some use cases I’m currently exploring with this setup:

• Data audits on existing Bubble apps to identify unused fields across multiple data types

• Letting non-technical team members (like sales or ops) query client data and run simple analysis

• Generating basic insights and graphs from app data without exporting anything

You can try the tool here:
https://bubble-mcp.com/

And if you want to understand how it works step by step, I wrote a full guide here

Print_[memory] — Digital Garden

What’s next

The next step is to allow adding Bubble backend workflow endpoints, not just Data API endpoints.
This would make it possible for AI to trigger custom business logic.

Another idea I want to explore is using ChatKit on top of this MCP server, so your Bubble app itself can expose a chat-based UI to interact with custom workflows.

I believe that more and more products will offer chat-based interfaces, not as replacements for UIs, but as an additional layer on top of existing apps.
In that sense, chat becomes both:

• a user-facing feature

• and an API-like interface for interacting with your product

All of this becomes possible thanks to MCP.

[code-escapee]

Sebastian, this is fantastic and is super helpful. Thanks!

[philipp.laengle107]

Awesome, thanks for sharing!!!

[ahmed.elkaffas]

This is awesome. I will check it out soon. Thanks for sharing !

[design.agx]

Thanks everybody!!

I’ll keep posting new features and tutorial on this!
But if you have already some use case, it would be great to hear how everybody else is planning to use it

[design.agx]

You can now add Backend workflows to your MCP and define the parameters you need to pass!

This is good for custom flows that involve subsequent actions!

This will help a lot to integrate chatkit later on, as we can send custom responses to fit the widgets that you can define in Chatkit

[design.agx]

Given that I have holidays this week, I’m profiting to work on the Chatkit integration!

Now you will be able to have multiple Chatkit interfaces that can use your MCP and are connected to OpenAI Agent Builder

I’ll be making tutorials next, after finishing this feature. Let me know your usecase or if you need some guidance, so I can write about it in the next tutorial!

[design.agx]

Update for the Chatkit Integration!

Added a feature for giving the AI context of the Current User, which will be the bridge for making sure the AI agent respects the privacy rules attached to that user when asking it to retrieve data trigger actions

Also, improve the logs view to have a better understanding of how the calls are being made to the MCP

Next step is to build a demo using all this to build an experience like https://chatkit.world/ where the user can interact both with the UI and Chatkit to perform actions in the app

[carlovsk.edits]

In the end, the logic behind MCP and the tools in Bubble.

Would it be something like this?

How To Build AI Agents with Multi-Step Tool Calling in Bubble 🛠️

A concise, modular blueprint you can adapt to any app. I call this a blueprint as it’s not a copy-paste but just gives some ideas that might help you get unstuck. Why this guide exists I just landed on a multi-step AI agent workflow that I’m happy with and wanted to share with you all as I haven’t seen many solid approaches for this. Bubble makes dealing with tool calls tricky for two primary reasons - JSON strings are hard to deal with in Bubble, and looping is hard in Bubble. This guide lay…

[design.agx]

It’s very similar, but there are a few advantages to using an MCP.

With MCP, the multi-step logic is no longer handled inside Bubble, so you don’t need backend workflows to loop and manage tool calls.
This is because, instead of defining the tools in Bubble, the tools are defined in the MCP itself.

Another benefit is that once you have an MCP, you can connect it to other clients and platforms like ChatGPT or Claude (as connectors), OpenAI’s Agent Builder, etc., without changing your Bubble logic.

[code-escapee]

and most importantly CLINE…

[antoine3]

@design.agx Honestly, this is a real achievement considering how closed the Bubble ecosystem is. This is exactly what this company should be building, instead of improving their front.

Good job man !

[design.agx]

Thanks @antoine3 !

I think Bubble has lots of priorities and is hard to allocate the resources, but hopefully they will keep improving!

I’m always open if they want to acquire my tool one day haha

[design.agx]

New updates:

Adding parameters to the endpoints

This will allow the AI to know which data the endpoint needs

Improved the query tester to accept parameters

Once you test the query, you can save the response from the endpoint as a MCP resource

Chatkit Integration

You can set up which backend workflow from Bubble should be triggered when a click in a widget happens
You can also define the payload that it will send to!

For later this month, I’m preparing a guide on how to set this all up in your Bubble app and build an experience around Chatkit.