Your repository may contain the app’s code, but the working product may still depend on services in another account.
An AI app builder’s code export is a starting point for portability. To move a working product, a founder also needs to account for data, sign-in, uploaded files, external integrations, and the new hosting environment. A successful download cannot verify all of those dependencies.
Lovable and Base44 both document GitHub integration, with different access conditions. Their documentation helps define what to inspect before committing a real business to a platform. It does not establish that a particular customer’s app will move without repair.
This is a documentation-based decision guide. MustHave has not migrated matching production applications for this comparison. The proposed rehearsal below is a way to evaluate your own project with evidence at each step.
Lovable separates code sync from a complete move
Lovable’s GitHub documentation describes two-way project-code sync. The standard github.com connection is available on all plans, while direct codebase download is a paid-plan feature. Enterprise GitHub connection types have their own plan conditions.
The same guide distinguishes workspace connections from a project’s repository link. That matters when a contractor sets up a project. A founder needs to know which account owns the repository and which workspace role can change the connection.
Lovable’s ownership guidance also separates the frontend from backend services. It describes retaining those services while moving the frontend, or migrating the broader stack. Newer TanStack Start projects and older React + Vite projects have different hosting needs.
A hosting tutorial written for a static frontend may therefore be the wrong starting point for a newer app. Inspect the actual project structure and runtime before choosing a destination. The builder’s name does not determine every project’s deployment model.
Base44’s GitHub access starts with the app owner
The Base44 integration guide says two-way GitHub sync requires Builder or higher. Only an app owner can connect or reconnect the app. The guide also distinguishes the current sync path from a legacy one-way export connection.
For a team, check repository ownership and app ownership together. A developer who can edit files may still be unable to reconnect a broken integration. Record the person who can resolve that problem before an urgent handover.
The GitHub guide establishes the sync route. It does not, on its own, prove that all runtime services can be replaced by deploying the repository. A founder should obtain the relevant service and data migration instructions before planning a complete move.
Define which kind of move you need
A code backup, a change of frontend host, and a complete departure are different projects. Each needs a different acceptance rule. Confusing them can turn a small deployment task into a bigger one after work has already started.
| Goal | Useful evidence | What it leaves open |
|---|---|---|
| Keep a code backup | A repository under an account you control | Whether the product runs elsewhere |
| Move the frontend | A working deployment at the new host | Dependence on the original backend |
| Move data and identity | Reviewed records and tested sign-in at the destination | Other integrations and operational jobs |
| Leave the original platform | Required features work without its services | Ongoing maintenance and replacement costs |
A founder moving only the frontend can deliberately retain a managed backend. That may reduce the amount of work. The important point is to describe the remaining dependency accurately when budgeting and planning recovery.
For a complete departure, define the product functions that must survive. A public homepage is often the easiest page to reproduce. A returning customer’s account and access to their purchased service provide a more demanding test.
Create a dependency record before touching production
List the service behind each critical product function. For every service, record the account owner, export method, and replacement decision. Keep credentials in the team’s secret-management system; the dependency list should identify them without exposing their values.
Include operations that happen outside a page load. A scheduled renewal, an email callback, or a background import can remain invisible during a visual review. Ask which events create or change records after the visitor has closed the browser.
Lovable’s external-hosting guide treats authentication providers, environment settings, and storage files as separate migration work. It also distinguishes database structures from the records inside them. A storage bucket definition cannot substitute for its uploaded files.
For each item, write how the team will verify success. Counting table rows helps detect a missing transfer, but it cannot establish that every relationship or permission is correct. Pair counts with a small set of known records and actual user actions.
A six-part rehearsal makes the risk inspectable
- Identify the repository revision and the intended destination.
- Build a separate environment without changing the live domain.
- Restore an authorized test copy of the required data.
- Configure test credentials and sign-in callbacks for that environment.
- Exercise the product functions in the acceptance record.
- Record a rollback path before planning the public switch.
Keep the source application available while examining the copy. Compare a known user’s records at both ends and verify which environment receives writes. Avoid letting a staging test send customer messages or charge a real payment method.
Test a new account as well as an existing one. A restored customer record can look correct while new registrations fail. A database import can preserve a row while the authentication system still needs separate handling.
Use the same discipline for file access. A visible filename does not prove the underlying file moved or that its permissions survived. Open a representative private file through the intended user account in the destination environment.
Budget for the move with a break-even calculation
Consider a hypothetical service bill of $80 per month on the current setup and $35 on a replacement. The gross saving is $45 monthly. If migration requires twelve hours valued at $40 per hour, the assumed setup cost is $480.
Ignoring other changes, $480 divided by $45 gives a break-even point of about 10.7 months. This is a calculation from illustrative inputs, not a price comparison between Lovable and Base44.
Now add one hour of extra monthly maintenance at the same $40 hourly value. The net saving falls to $5 per month, and the same setup cost takes 96 months to recover. A small infrastructure saving can disappear when operational work moves onto the founder.
The move can still make sense for control, a required integration, or a customer’s hosting conditions. Give those benefits their own acceptance rule. Avoid presenting an operational choice as a guaranteed cost saving.
Ask for evidence during a developer handover
Request the repository, deployment instructions, and a dependency record with named account owners. Ask the developer to show the product working in the environment the company controls. The demonstration should include a critical write operation and a recovery path.
A useful handover also explains what remains on the original builder. That prevents a later cancellation from disabling a service the team thought it had moved. Change subscriptions only after the relevant replacement functions pass review.
Explore the existing Lovable listing for product context, and the automation cost guide for a way to include review labor. Which dependency would be hardest for your team to replace if you had to move tomorrow?



