Hey everyone 
I’ve been building with Bubble for quite a while, and one thing that kept bothering me was how much time it can take to build different AI features into an app.
You need a chat → find a plugin or handle the integration yourself, then build the workflows and logic.
Need image analysis → new workflows, API calls, logic.
Need text generation → Pretty much the same process again.
Need voice → Same thing — new API calls, workflows, and more pieces to connect together.
Then you want Claude instead of OpenAI, or Gemini for a specific use case, and suddenly you’re changing plugins, workflows or API logic again.
So I decided to build something that I think could help you save a lot of time.:
AI Toolkit — Chat, Voice & Vision
The idea is pretty simple:
Bring the most commonly used AI features into one Bubble plugin, so you can add whatever AI functionality your app needs without having to build your AI stack around a different tool every time.
It currently supports OpenAI, Anthropic and Google Gemini through the same interface.
So switching providers is basically a dropdown instead of rebuilding your implementation.
What’s inside?
There are currently 5 visual elements + 21 server-side actions.
AI Chat
A complete streaming chat component with:
- Live streaming
- Markdown
- Light/dark themes
- Suggested prompts
- Stop & regenerate
- Conversation history
- Saving/restoring conversations from Bubble’s database
It can also work as an AI agent.
You can give it tools and let the AI trigger normal Bubble workflows, receive the result, and continue the conversation.
AI Voice
Real-time voice conversations using OpenAI’s Realtime API.
The user speaks → the AI responds with voice → the user can interrupt it and continue naturally.
You also get live transcription of both sides.
One of the things I paid a lot of attention to here is security: the API key never needs to be exposed in the browser.
The plugin creates a short-lived token server-side and uses that for the realtime session.
And because the voice assistant can use Bubble workflows as tools, it can actually do things during the conversation.
For example:
“Do you have any appointments available tomorrow?”
The assistant can query your Bubble database, find an available slot, and respond with the result.
AI Text
Same AI engine, but without forcing you into a predefined UI.
This is useful when you want to build your own interface and just need AI generation inside it.
It also runs client-side, which makes it useful for interactive experiences where you don’t want to depend on Bubble’s server-side 30-second action limit.
AI Autocomplete
Suggestions while the user is typing, similar to the autocomplete experience you see in code editors.
The interesting part is that the suggestions can use the actual data from the record being edited.
For example, instead of giving generic suggestions, an AI field could know that the current property is:
87m² · 3 bedrooms · renovated in 2023
and generate suggestions based on that actual record.
⌘ AI Command
A command palette for your Bubble app.
Users can type something like:
“Give Nova Solutions 20% off their next order”
The AI interprets the request and returns the command + arguments that your Bubble workflows can execute.
And the server-side actions?
There are 21 actions covering things like:
Text
- Generate Text
- Transform Text
- Universal AI Request
- Get JSON Value
Images & documents
- Analyze Image
- Analyze Document
- Generate Image
- Edit Image
Voice
- Transcribe Audio
- Text to Speech
- Create Voice Session
Knowledge / RAG
- Create Embedding
- Compare Texts
- Index Text
- Search Knowledge
- Ask Knowledge
- Search Web
Memory & data
- Remember
- Recall
- Moderate
- Analyze Data
The goal wasn’t to create 50 different AI actions just for the sake of having a huge list.
The idea was to cover the things that actually come up when you’re building an AI-powered Bubble application.
The plugin is still new, so I’m sharing it here partly to get it in front of more Bubble developers and partly because I’d genuinely like feedback from people building products with Bubble.
There is also full documentation with a practical example for every element and action.
AI Toolkit — Chat, Voice & Vision
Would love to hear what you think, especially if there’s an AI feature you use frequently in Bubble that you think belongs in a toolkit like this. 