Running a migration end to end

Buildprint Migration is a powerful feature to allow both automatic codegen, and AI agents, to migrate your Bubble app to code, effectively, reliably, and securely.

This guide details how you can run this in practice, after receiving access to the migration feature.

Preparation

There are a few things you can do in your Bubble app to make sure the migration is as straight forward as possible before beginning:

• make backend workflows non-public if they don't need to be public

• delete unused pages/reusable elements/features

• delete unused data types, fields, and option sets/values

Essentially, trim the fat, and make your application as simple as possible.

Create your project tech stack accounts

Create accounts for the services your migration will use. We will have advised you on the right option for you, but you can choose any of Buildprint's supported options. These might include:

• Vercel

• Github

• Supabase

• Trigger.dev

• Convex

Ask Buildprint to create a migration for your project

All migration execution happens via Buildprint CLI. It can be monitored using the dashboard on Buildprint web (or by asking your agent).

It is strongly advised to use the latest SOTA model for leading the migration. Right now, that would be Claude Fable 5.1. It may cost you more (either in Claude Subscriptions or API billing) but will yield the best results.

Ask your agent to create a migration for your project. You'll then see the migration appear in your Buildprint dashboard in your project.

Complete setup

Your agent will be able to complete setup. At any time, it may ask you questions (e.g which tech stack to choose). You can also ask for its advice here.

It will ensure everything is in place for a full migration.

Run the migration

A migration is broken down into a series of smaller tasks. These tasks are generated by Buildprint. Roughly speaking, a migration follows this process:

1. Complete setup

2. Buildprint runs automatic task generation and code generation

3. An orchestrating agent distributes tasks to implementation agents, in parallel where possible

4. The same orchestrating agent distributes reviews to different agents, so they can review each other's work

5. Once complete, the CLI is able to take a data + file backup of your Bubble app, and upsert it to your new database, as well as manage end to end deployments.

To kick off the orchestrating agent, ask it to read the 'orchestrator agent orientation for the migration feature'. We recommend avoiding using loops/goals (whatever feature is in your relevant agent harness) as it will often prod the agent for input even while subagents are working.