Research Note

AI App Builder Portability Research Note

Code access is one part of portability. A usable exit also requires the application data, schema, files, identity records, secrets inventory, scheduled work, integration

Aug 4, 20262 min readBy Dalton Anderson
In this article

AI App Builder Portability Research Note

Code access is one part of portability. A usable exit also requires the application data, schema, files, identity records, secrets inventory, scheduled work, integration contracts, deployment configuration, observability, dependency licenses, and operational knowledge needed to run the system elsewhere.

Base44's current GitHub documentation describes two-way sync on eligible plans. It also says the connection is permanent, cannot be transferred back to Base44, changes are synchronized automatically, and pre-connection version history is not available through the repository. Its quick-start documentation says ZIP and GitHub export require the Builder plan or higher.

Emergent's GitHub documentation describes push and pull workflows, branches, collaboration, and recovery through GitHub. That establishes documented code movement. It does not by itself establish a complete export of hosted databases, deployments, accounts, or external-service state.

Lovable's GitHub documentation describes two-way synchronization and treats GitHub as the source of truth after connection. Its deployment and ownership guidance distinguishes Lovable Cloud, hybrid, and self-managed setups. It states that generated applications can be hosted elsewhere while the Lovable editor and agent remain managed services.

Replit's version-control documentation describes Git, GitHub import and push, branches, and Agent checkpoints. Replit's publishing documentation explains that published apps run from snapshots and should not rely on the deployment filesystem for persistent data. Code movement therefore does not automatically move Replit-managed databases, authentication choices, stored secrets, or publishing configuration.

Firebase Studio allowed code view and GitHub-based deployment routes. Google now states that new App Prototyping agent workspaces were disabled on June 22, 2026 and recommends migration to Google AI Studio. Historical code access does not make the retired interface a current procurement option.

Exit test

The practical test is not whether an export button exists. It is whether a different team can restore a sanitized copy in a clean account, recreate infrastructure from documented configuration, import representative data, authenticate test users, run critical workflows, observe failures, roll back a release, and explain remaining platform dependencies.

A product can be portable enough for a prototype while remaining expensive to move after users, data, automations, and managed services accumulate. The buyer guide should ask for evidence at the project's expected future scale, not only at day one.

Sources

Follow the evidence.

  1. docs.base44.com: githubdocs.base44.com
  2. web.dev: vitalsweb.dev
  3. docs.replit.com: replit appsdocs.replit.com
  4. docs.replit.com: build with agentdocs.replit.com
  5. csrc.nist.gov: finalcsrc.nist.gov
  6. help.emergent.sh: 272715 features and toolshelp.emergent.sh
  7. firebase.google.com: migrating projectfirebase.google.com
  8. owasp.org: www project application security verification standardowasp.org
  9. docs.base44.com: Quick start guidedocs.base44.com
  10. w3.org: WCAG22w3.org
  11. firebase.google.com: get started aifirebase.google.com
  12. help.emergent.sh: plans and creditshelp.emergent.sh
  13. docs.lovable.dev: githubdocs.lovable.dev
  14. docs.lovable.dev: getting starteddocs.lovable.dev
  15. firebase.google.com: overviewfirebase.google.com

From this episode

Two useful next steps.

Evergreen · 1 min

What One-Shot App Generation Actually Proves

A one-shot AI app build can prove initial instruction-following and visible interaction. It cannot prove security, correctness, scale, or demand.

Research Note · 1 min

Vibe Coding Benchmark Method Research Note

A useful AI app-builder benchmark must answer a decision rather than manufacture a universal leaderboard. The decision might be which tool best supports a team's internal

Return to the episode
AI App Builder Portability Research Note