Hi everyone,
This is the October community update. You can read last month’s update here.
I’ve been talking about how one of our biggest focuses has been bringing Bubble fully into the age of AI. That meant making sure our Agent consistently delivers the magical app building experience people get from tools like Claude Code, and opening Bubble up to every agent through an MCP.
Today, we’re shipping both: Agent 2 is out of beta and available to everyone, and the Bubble MCP is live in beta. You can read the full announcement here, and the key changes are also summarized below.
I’ll be honest — it’s been a tough year for the Bubble community. It’s a hard sell when AI can deliver instant gratification, and the costs of vibe coding only become apparent down the line. We believe there’s a best of both worlds: the speed of AI combined with the control you get when you can understand, refine, and scale exactly what your app does. But, we haven’t had the AI capabilities to back that vision up. Now we do. With this release, we are adding two strong options for building in AI with Bubble:
-
Use our Agent when you want a batteries-included build experience. We’ve stress-tested it against thousands of builds, and we’re confident that it provides a balance of speed and quality that stacks up against any AI builder on the market. If you tried it a month ago and it didn’t meet your expectations, please test again: we’ve made some big changes.
-
Use our MCP (beta) when you want full control of the AI build process. It lets you bring your own coding agent, skills, workflows, and components, tailored to enable you to leverage your existing Bubble expertise with complete freedom
With that, let’s go deeper into what we released.
What we launched: Agent 2, Bubble MCP, and an updated pricing structure
Agent 2 is available to everyone as of today
We’ve been building the Agent in public for about a year and it’s now officially available on all web and mobile apps. Thank you to everyone in the beta who told us what was and wasn’t working. A lot of Agent 2 came directly from that feedback.
Here’s what’s new in Agent 2:
-
It’s now 1.5X better at building exactly what the user asked and 2X better at ensuring the result is fully functional, as measured by our internal eval suites. That makes the difference between having to look over its shoulder vs trusting it to take on complex builds and edits, like adding a full feature to a larger, established app. For advanced users, like readers of this forum, it’s more of a true build partner.
-
It’s also significantly better at individual tasks like creating designs, writing expressions, setting up data, and working with vague or unclear prompts.
-
It has an improved UX. Preview is now its own tab inside the editor so you can see what your app looks like right away, and the Agent stays pinned as you move between tabs like data and design. You can also ask it to make changes without leaving preview.
We’re updating pricing to align with usage
Starting today, iterating with the Agent uses AI credits, which come included in every plan. We run the Agent on the latest models, and credits let us keep it up-to-date as models improve.
Credit-based usage models are common these days, but unlike most AI app building platforms, credits aren’t the only way to build on Bubble. You can make edits in the Bubble visual editor or you can get started with the new Bubble MCP. You’re always in control of your AI costs, and building is never dependent on AI credits.
Lastly on the credit front, as a thank you to our community, apps created on or before October 6, 2026 will receive double the regular credit amount through October 31, 2026. Standard credit amounts will begin November 1, 2026.
We’re also changing the cost of the web-only Starter plan (no other plan’s pricing is affected) starting October 20, 2026. Read more about these updates on the blog here.
The Bubble MCP is live in beta
If you’d like to build with your own agent, like Claude Code or Codex, now you can! Use the MCP to hook in your own agent. There are no request limits, even on the Free plan. It also doesn’t use your Bubble AI credits, though you may run into usage limits from your provider.
The MCP is built for experienced builders who want to edit and grow apps they already have by connecting to LLMs. If you want to build with AI without any custom setup, the Agent is a better fit.
Learn how to connect the MCP server here. Once it’s connected you can:
- Look up and explore your app’s data without opening the editor
- Clean up unused workflows or data fields
- Build and edit UI and features in your app
Other features, such as reading logs, merging branches, and deploying, are still in progress.
Both the Agent and MCP are live today! Give them a try and leave us feedback in the thread below — we read it all very closely. If you have questions about the Agent or MCP releases, join us for a special edition of Office Hours. We’re running a few sessions a week until Nov 6 — pick a slot that works for you. The RSVP link will expire on November 6. Save your spot →.
For a full list of what else we shipped this month, check out bubble.io/releases
What we’re working on
-
Workflow engine: We’re currently in the alpha testing stage of the workflow engine revamp. This update is a total overhaul of how our platform handles workflows and will enable branching and looping. Big thank you to our alpha testers for their feedback on this. We’re targeting moving this feature into beta toward the end of the year.
-
Continue improving MCP: Now that the MCP is in beta, we’re working on a number of follow-ons, feedback from the community, and more per-app permission configurations.
-
App data tab: We’re redesigning the app data tab! Thank you to our current beta testers, including @heythere, @georgecollier, and others. The big unlock with this update is inline editing, so you’ll be able to edit each cell in your data tab just like you would in a spreadsheet. Right now we’re implementing beta tester feedback and doing additional design QA, and targeting getting this out in November.
ICYMI
-
The team hosted a Launch Lab fireside chat, “Fix Your Pitch, Then Fix Your Reach,” on messaging and distribution tactics for builders getting ready to launch. You can watch the recap here.
-
As a follow-up to that session, we published a landing page message auditor. You can use it to audit a live landing page against a reader-first messaging framework at launch.bubble.io. (Note that it only works if your landing page is built on Bubble!)
-
Emmanuel hosted a really great AMA and demo session. He gave an early look at the MCP and answered a lot of your questions. Check out the replay here.
New hires
In keeping with our busy month, we also had a big batch of new hires join the team! Big welcome to:
- Meg, director of product design
- Ryan, senior product designer
- Alex, senior fullstack engineer (mobile)
- Denise, senior commercial success manager
- Varsha, customer success manager
- Nicole, technical product support specialist
That’s all for this month!
— Josh and Emmanuel
but i’m sure you have your reasons not to…
