Leaving Bubble means rebuilding, not converting: Bubble does not export runnable source code, so you re-create the app in a stack you own (for example Next.js and React on the front end, a Node or Python API, and Postgres) while copying the data out through the Data API or CSV. Run Bubble and the new app side by side and move one feature, and one data type's system of record, at a time. Success depends on the inventory: every workflow, conditional and privacy rule you fail to write down is behavior your users lose without warning.
Two facts drive most Bubble exits: how Bubble charges, and what you can take with you.
On pricing, Bubble announced on April 6, 2023 that a usage-based metric called workload would replace server capacity, effective May 1, 2023. Workload is measured in workload units (WU), and every search, workflow run and API call spends them. The pricing page lists web plans billed annually at Free ($0, 50K WU), Starter ($32 per month, 175K WU), Growth ($119, 250K WU), Team ($349, 500K WU) and custom Enterprise; free plans cannot enable overages. Bubble's own November 6, 2026 cost post gives the billed-monthly web prices as Starter $39, Growth $134 and Team $399, with separate Mobile and Web+Mobile tiers. Past your allowance, the default overage is $0.30 per 1,000 workload units.
On portability, Bubble's support article is direct: "We do not support on-premise self-hosting or code exports at this time." The JSON export only imports into another Bubble app. A newer article, Exporting your Bubble app, repeats that Bubble does not export ready-to-run code and points to the Bubble MCP (in beta) for an AI-assisted rebuild, noting that the result still needs to be built, reviewed and tested like any other software project.
The business consequence: your cost grows with user activity, metered in a unit you cannot tune with the usual levers of your own stack, such as indexes, caching and query plans. Your product's logic lives in a visual editor that only Bubble can run, which comes up in security reviews, enterprise procurement questionnaires and acquisition diligence. And the engineers you hire to scale the product want a codebase, tests and CI, not an editor.
A Bubble app has no repository to read, so the inventory is the specification: one row per item, an owner per area.
Here the strangler fig works by feature and by data type. Bubble stays the system of record for everything until a feature moves, and each data type has exactly one writer at any moment.
Pick the stack and design the schema before writing screens. A common target is Next.js on React (React 19.3.0, September 9, 2026, is the current release line; Next.js is on major version 16, with 16.4 released October 6, 2026), an API in Node (Fastify or NestJS) or Python (FastAPI or Django), and Postgres. Translate each Bubble type into a table with a bubble_id column holding the Bubble unique ID, so every reference can be resolved during sync and every row traced back. List-of-things fields become join tables. Option sets become lookup tables. Decide now which Bubble quirks to keep and which to fix. Exit criteria: schema reviewed by the tech lead and the Bubble owner, and a sample export loads with every reference resolved.
Enable the Data API for each type you need and build a sync job that pages through each type and upserts into Postgres by bubble_id. Use a CSV export from the Data tab for the first load if faster, then incremental pulls filtered on Modified Date. Copy files from Bubble's storage into your own object storage (S3 or equivalent) and keep a mapping from old URL to new URL. Budget the workload units: the sync itself runs searches against Bubble and spends WU, so watch the workload chart after the first full pull and throttle if overage starts. Exit criteria: scheduled sync with per-type reconciliation matching, and no orphan references.
Pick an auth layer you control: Auth.js, Clerk, Auth0, Amazon Cognito or your framework's own. Import users by email with their bubble_id and role. Plan for password resets from the start rather than trying to carry password hashes across; the cleanest pattern is a "set your password" or magic-link email the first time a user signs in to the new app, with SSO users re-linked by verified email. Turn the privacy-rule inventory into authorization code: one policy function per data type in the API layer (optionally backed by Postgres row-level security), and a test per role asserting it sees exactly the records and fields it saw in Bubble. Exit criteria: test users for every role sign in, and the per-role visibility tests pass against synced production data.
Start with read-only features (admin dashboards, reports, exports, search) that run against the synced copy and cannot corrupt Bubble data. Then move features that write, one at a time. Each Bubble workflow becomes service code with a test; backend and scheduled workflows become jobs on a queue (BullMQ for Node, Celery or RQ for Python) and cron schedules; database trigger events become explicit calls in the service that changes the data. Replace plugins with the underlying library or API directly (the Stripe SDK instead of a Stripe plugin), and API Connector calls with typed clients. When a feature moves, its data type's system of record moves with it, and the new app pushes changes back to Bubble through the Data API (or calls a Bubble backend workflow exposed as an endpoint) so screens still on Bubble see current data. Exit criteria per feature: parity tests pass, a flagged cohort used it without regressions, and write-back works.
Once most daily use is on the new app, move the last data types. Freeze writes in Bubble (hide edit controls with a condition, or put pages behind a maintenance notice), run a final delta sync, reverse the sync direction so Bubble becomes a read-only mirror, and send the remaining users to the new app. Exit criteria: no writes to Bubble for an agreed period, and support tickets at or below baseline.
Set up 301 redirects from old Bubble page URLs to new routes, rewrite any stored file URLs to your storage, take a final CSV export of every type as an archive, and only then downgrade or cancel the Bubble plan. Remember the JSON export only restores into Bubble, so it is a backup, not an archive you can query. Exit criteria: no traffic or API calls reaching Bubble, files served from your storage, and the archive tested by loading it.
Users move by feature flag and by route, not all at once. Put the new app on its own hostname or behind a reverse proxy (CloudFront, Cloudflare or an Nginx edge) that sends migrated paths to the new app and everything else to Bubble. Gate each feature with a flag (LaunchDarkly, Unleash or a users-table column) and widen by cohort: staff, friendly customers, everyone. If you change the primary domain at the end, lower the DNS TTL a couple of days before the switch.
For write features, run a shadow period before the flag opens: the new service processes the same inputs Bubble did (from the sync or a copy of the webhook) without committing, and a comparison job diffs its intended changes against what Bubble actually wrote. Differences point at a missing conditional or rule. Never let both apps write the same data type for real; that is how records diverge and stay diverged.
Go/no-go checklist for each feature:
Rollback path per phase:
A Bubble rebuild is product engineering more than infrastructure. A typical shape: a tech lead who owns architecture, data model and cutover; two senior full-stack engineers on React and Next.js plus the API language (see Next.js developers and Node.js developers, or Python developers for a FastAPI or Django backend); one engineer focused on the data pipeline, sync and reconciliation; QA to turn the inventory into parity tests; and a designer if the UI is being rethought rather than copied. The person who built the Bubble app is part of the team in practice, even part-time. Many teams begin with a pod of 3 to 5 senior engineers (a tech lead plus engineers), and a full pod is about seven.
Ryz engineers on your team start from $7,000 per month (our published range is $7,000–$15,000 per engineer per month, mostly mid-level to senior), quoted per team. Project cost = team size × months × monthly rate; the months depend on how many workflows and privacy rules the inventory finds, how many plugins need rebuilding, and how much of the UI you redesign. Check React developer rates for the seniority breakdown. Ryz engineers work on your team, reporting to your leads. Once you say go, engineers can typically start in under two weeks. Scope your team, or see how our SaaS development teams build the product you are moving to.
A Bubble app is a large, mostly mechanical specification, which suits AI-assisted work. Our AI pod uses coding assistants to draft the Postgres schema and migrations from the data export, generate TypeScript or Python types and API clients for each type, turn plain-English workflow rows from the inventory into service stubs with failing tests, and write the reconciliation queries that compare Bubble and Postgres. It also triages shadow-comparison diffs, grouping hundreds of differences into a handful of missing rules. Bubble's own Bubble MCP (beta) can produce a starting rebuild in code; Bubble says it does not guarantee perfect fidelity, so treat its output the same way as any other draft.
Where it does not help: deciding which Bubble behaviors are features and which are accidents, writing authorization policy from privacy rules, and choosing how users move through password reset. Every generated schema, policy and workflow port goes through human review and has to pass the parity tests recorded from Bubble before it ships.
Ryz is not the right fit if you want a self-serve freelance marketplace or hourly short gigs to patch a Bubble app; our teams do the rebuild.
It depends on what the inventory finds: the number of data types, workflows and privacy rules, how many plugins encode logic you must rebuild, how much of the UI you redesign rather than copy, and how quickly users can be moved through the password reset. Size the work after the inventory.
No. Bubble does not export runnable source code or support self-hosting, and its JSON export only imports into another Bubble app. You can export your data through the Data API or CSV, and Bubble's MCP beta can draft a rebuild, but the application has to be rebuilt, reviewed and tested as new software.
Yes, and that is the safest path. Bubble stays the system of record while a sync copies data into Postgres, and each feature moves to the new app behind a flag. When a feature moves, its data type gets a single writer, and changes are written back to Bubble so remaining Bubble screens stay accurate.
Plan for it. Build the migration so it does not depend on carrying password hashes over, and make the first sign-in on the new app a set-your-password or magic-link step. Users on social login or SSO can usually be re-linked by their verified email.
Engineering cost is team size multiplied by months multiplied by the monthly rate per engineer. Add the overlap: you keep paying for the Bubble plan, plus any workload overage from the sync, while both apps run, and you take on hosting and monitoring for the new stack. The inventory drives the months, so price the project after it is done.
Questions we didn't answer? Email info@ryzlabs.com.
Thanks — your message has been sent. We’ll get back to you soon.
Something went wrong while sending your message. Please try again or email info@ryzlabs.com.