Skip to main content
Almost an Architect
Current page: Privacy

Privacy

What we collect, what we don't, what we do with it.

The short version

We don't have user accounts yet, so there's very little to collect. Sessions are stored in your browser's localStorage and in our Redis cache for 24 hours; we don't link them to any persistent identity.

What we store

  • Session data (your chat with an architect, the brief, the current plan) — stored in Redis with a 24-hour TTL. After 24 hours of inactivity, deleted.
  • Uploaded reference images (when you provide them) — stored in Vercel Blob. Used by the AI to generate exterior images. Deleted with the session.
  • Generated images — same lifecycle as session.
  • Usage metrics (token counts, model costs, image-generation calls) — for our cost monitoring. Aggregated; not tied to individual sessions in a way we use.

What we don't collect

  • Real names, email addresses, phone numbers (no accounts yet).
  • Browsing history outside the app.
  • Tracking across third-party sites.
  • Your floor plans for training. Plans you generate are yours; we don't use them to train models.

AI provider data sharing

Chat messages are sent to Anthropic Claude (the model that powers the architect personas). Image-generation prompts are sent to OpenAI gpt-image-1 with dall-e-3 as fallback. Both providers receive your chat content under their respective data-handling policies. They process the request and return the response; we don't share your sessions with them for training.

When accounts ship

This page will expand to cover email addresses, project storage, billing data, and the standard account-related telemetry. You'll be notified of the changes through the app and asked to acknowledge before continuing.

Last updated: 2026-05. Questions? See the contact page.