additional feedback is here~!
unsubscribedafter subscribing starter plan.log out button and I see sign out button in the setting tab.Iām waiting to built with autobubble. Please check the issues. Thank you carlovsk.
Hey! Thanks for the heads up, and sorry for the friction.
Honest confession: Iāve been so obsessed with making AutoBubble actually ship good plugins that I kinda left Stripe alone in a corner. It got jealous, started a race condition, and locked a couple of users (including you 
) on the ātrialā plan even after paying. Drama. Fixed now.
Your account is now active on the starter plan.
On the sign out: it does exist today under Settings ā Sign out, but youāre right that it should be in a more obvious spot (header / account menu). Adding it to the list.
Thanks for being patient and for the feedback, genuinely helps. Excited to see what you build! 
It seems thereās a bug where, when running multiple chats simultaneously, the Agent reflects the results into other plugins. It appears the Agent is not executing in parallel.
Specifically, I created four plugins and am chatting with the Agent in each, but the results from chat A are being applied to plugin B. This ended up overwriting existing work. Even the titles ended up being synchronized..!
I like Autobubble but this should be fixed ASAP..!
Hey!
Thank you so much for all the feedback, you have no idea how much it helps me. I canāt test all the scenarios, and this public beta has been great thanks to your feedback.
Iām already fixing it and Iāll be back with updates later!
I would really appreciate it if you could also look into the following:
The reason I am confident in Autobubbleās success is that it enabled me to quickly and easily develop two highly useful plugins.
The first is a plugin that accepts any type of field (including option sets), performs arithmetic operations such as A+B+C/D*2, and returns the calculated result.
The second is a plugin that interacts with a Vimeo player embedded via an HTML element. By simply providing the elementās id attribute, it can return playback time, total duration, and play status in real timeāwithout relying on ādo when condition is trueā events or ārun JavaScriptā actions.
I was able to build both of these in just one day because they address problems I had long wanted to solve more efficiently. In other words, these were features that made me think, āI chose Bubble to build things quickly and easily, but implementing this within Bubble is unnecessarily cumbersomeāhow can I simplify it?ā
For Bubblers like myself who donāt have coding experience but still want to solve practical problems, this tool is extremely valuable.
Thanks for the detailed feedback, this is exactly the kind of input
that shapes the roadmap. Quick reactions to each point:
Highest priority for me. Two fixes coming:
Agreed, Plan Mode is on the roadmap. The idea is to let you scope and
validate the approach with a lighter model before committing credits
to the actual build step.
Also on the roadmap. Looking at OAuth-based integration with the
Claude subscription so Max users can plug their own account in and
skip the double billing.
Hey everyone 
A few updates were released earlier this week:
Connect with your Claude account (no API key needed) You can now sign in with your Claude Pro or Max subscription directly, no need to generate an API key or manage billing on the Anthropic side. If you already pay for Claude, you can just plug your account into AutoBubble and go. API key support is still there for anyone who prefers it.
No credits charged on failed requests If a generation fails or the AI doesnāt return a valid patch, you wonāt be charged credits anymore. Only successful edits consume credits.
Bug fixes A handful of fixes around chat isolation per plugin and the AI provider picker, plus smaller polish across the editor.
Whatās next: Plan Mode The next feature Iām shipping is Plan Mode, where you can work with the AI to draft a plan first, review and adjust it, and then have it implemented in your plugin. Should make bigger changes a lot more predictable.
As always, feedback is very welcome, Iām right here in the thread.
If you have used AI to build Bubble plugins, you know the feeling: you type a prompt, hit build, burn credits, and only then find out the AI touched the wrong element, missed a state, or built something slightly off from what you had in your head. You either accept it or pay again to fix it.
That is the exact problem we set out to kill with Plan Mode.
Plan Mode adds a review step before the build. Instead of going straight from prompt to spent credits, AutoBubble first drafts a clear, structured plan of everything it is about to do. You read it, refine it, and only build when it matches what you actually want.
Here is how it works.
You pick Plan Mode right from the chat, the same place you already work, and send your request.
Instead of a vague summary, you get a SPEC style plan with five clear sections:
Summary of what will change
Goal in user-story form
Affected Surface listing every element, action, field, state, event, and code path it will touch (by name and id)
Acceptance Criteria as a checklist of behavior you can verify in your running plugin
Steps, numbered and specific
You can refine it as many times as you want, edit it by hand, and the plan keeps its structure every time. Nothing gets built until you say so.
No more surprise builds. You see the blast radius before you spend.
Fewer wasted credits, because you catch wrong assumptions in the plan, not in the output.
Confident iteration. Refine the plan freely instead of re-running expensive builds.
It is live now. Give it a try on your next plugin and let me know what you think.
Havenāt tried this yet but looking forward to at some point 
What makes AutoBubble different from Buildprint by George?
Itās simple: Buildprint helps you build your app; AutoBubble creates and edits plugins for Bubble, for you to use in your app. 
