What we store, and what we send elsewhere.

What we collect

• Your account — email address, and a name if you give one. A password hash, unless you sign in with Google, in which case we store no password.
• Your projects — the prompts you write, the conversation with the builder, and every file it produces, including previous versions.
• Usage — tokens and estimated spend per month, so a plan can be metered. Not keystrokes, not mouse movement, not session recordings.
• Credentials you connect — provider keys, GitHub and Supabase tokens, and project environment variables. Encrypted at rest; see Security.
• Two cookies — a session cookie, and one holding whether you have asked to reduce motion. No advertising or cross-site tracking cookies.

What we send to a model provider

This is the part worth reading twice. To build a site, your prompt and the relevant contents of your project are sent to a third-party model provider over the network. That is how the product works and there is no version of it that does not do this.

Two things follow. Do not paste a credential into the chat — use the environment variable settings, which are encrypted and never sent to the model. And if you need model inference to happen under your own commercial terms, add your own provider key: the request then bills to your account and is governed by your agreement with them rather than ours.

Who can read your conversations

Operators of this instance can. Rather than imply otherwise: access is by named capability, the supportrole cannot read a conversation at all, and for the roles that can, a read requires a stated reason and writes an audit row before returning anything. If you want to know whether anybody has read a particular project’s conversation, ask — it is a query, and we will answer it.

We do not use your prompts or your generated code to train anything, and we do not sell or rent any of it.

Analytics

Measured on the server, not in your browser. There is no analytics bundle on any page — the content security policy would refuse one — so what we record is a short list of moments the server already knows with certainty: an account was created, a build finished, a site was published, a plan activated. No scroll depth, no heatmaps, no replays.

Payments

Handled by Razorpay on their own hosted page. Card details never touch this origin and we never see them; we store a subscription id, its status, and the date the current period ends.

How long we keep things

• Projects and conversations — until you delete them, or until you ask us to delete your account.
• Usage records — kept as billing history after a project is gone, because an invoice has to be explicable.
• Operator audit log— kept, and deliberately outlives the operator’s own account. A record of who read what must survive removing the person who did it, or removing them erases the trail.

Getting your data out, or deleting it

You never need to ask us for an export: every project can be pushed to your own GitHub repository in one commit, assets included, and the database is already in your own Supabase organisation. That is not a data-subject-access process, it is a button, and it is the same one the product is built around.

For account deletion, ask through the contact page. We remove your account, your projects, your files and your conversations. Billing records and the operator audit log survive for the reasons above.

Changes

If this policy changes in a way that affects what we do with something you have already given us, we will tell account holders rather than quietly updating the date at the top of the page.

Contact

Anything about this policy, or about your data: get in touch. A real person reads it.