Firebase Studio

Do not adopt. Firebase Studio was a pleasant free cloud IDE with a prototyping agent, and none of that matters, because the service shuts down on March 22, 2027 and everything built inside it will need a new home. The daily trait that once decided it, zero install and a VM that followed you between laptops, is now the reason you have to leave.

If you already have a workspace, spend the afternoon exporting it rather than extending it. If you are choosing today, pick Lovable for the hosted experience, or Bolt if you want the code in your own repo from the first prompt.

The risk is the pivot, already executed. New signups and workspace creation were disabled on June 22, 2026, which means the product can only shrink, and a product that can only shrink gets fixes last. While it lived, Agent (Auto-run) mode executed tools without approval inside a Google Cloud VM, the right place for that risk, and the docs said so plainly.

The consequence for anyone still inside is that bug reports filed now compete with a shutdown checklist for attention. Nothing changes this verdict; the decision was made upstream. What it does right: the shutdown notice came with a documented migration path rather than a redirect.

The migration document is precise. 1. Export is a Zip and Download button or npx firebase-tools@latest studio:export PATH, so the artifact is the file tree, reproducible from a shell. 2. The prototyping agent built Next.js applications from text, images or sketches. 3. Agent chat history is not part of the exported zip, though the files sit in ~/.idx/ai.

The consequence: the code survives migration intact and the reasoning that produced it does not, so anyone relying on the conversation as documentation should copy that directory by hand. The observation: the vendor kept the code and dropped the conversation, the correct priority.

This was Project IDX, relaunched as Firebase Studio in April 2025, and sunset on March 19, 2026, eleven months later. Two names and a shutdown in under two years is the pattern of a product nobody inside owned, and inside a company Google's size, ownership is the only runway that matters. Google is consolidating: AI Studio takes browser prototyping, Antigravity takes code-first work.

For buyers the lesson is about the parent, not the product: a free tool from a large vendor has no revenue line to defend it, so it lives exactly as long as the reorg allows. No acquirer, because nothing is left to acquire. Position: zero.

The demo was a browser IDE that needed no install. Procurement never had a line item: free, limited to three workspaces or thirty for Google Developer Program members, no seat, no SSO story, no contract, and no CI hook. The retention policy is now the sunset notice: all remaining data is permanently deleted.

That last line is the one that matters for anyone with a prototype still inside, because permanent deletion on a fixed date is a compliance event, not a feature. There is nothing to approve and nothing to renew. Not yet, and there will not be a later.

I never owned any of it: a Google VM I could not take home, Gemini only, no key field, no local model, an MCP client as the one open protocol, and a config I could not check into my own repo. When a hosted IDE closes, the shell history and the extensions go with it, which is the whole argument against hosted IDEs made in one announcement.

What survives is the Firebase side, Firestore, Authentication and App Hosting keep running, because that is where the vendor's meter lives. The editor was the free sample. The backend was the sale. If you must build on it, keep the backend swappable.