FreeLoom uses these client-side storage mechanisms:
| What | Purpose | Category |
|---|---|---|
| sb-...-auth-token (cookie) | Keeps you signed in between visits. | Strictly necessary |
| freeloom-current-student-id (local storage) | Remembers which student profile you last had selected. | Functional |
| freeloom-cookie-consent (local storage) | Remembers your cookie/AI-model choices so we don't ask every visit. | Strictly necessary |
| AI model weights (browser cache / IndexedDB) | Benny's AI model (Llama 3.2 1B, or Qwen2.5 0.5B on mobile), downloaded once and reused across visits so it doesn't re-download every time. Only starts if you accept this in the cookie banner or turn on Benny in Settings. | AI model (opt-in) |
We don't currently use analytics, advertising, or any third-party tracking cookies. The cookie banner you may have seen has Analytics and Marketing categories in it anyway, reserved in case that changes in the future -- if it ever does, this page and that banner will both be updated, and you'll be asked again before anything new is set. You can revisit your choice any time from the Cookie preferences link in the footer, or in Settings > About if you're signed in.
Benny's assistant-mode chat and activity-drafting help run entirely on-device, using your browser's own graphics hardware (WebGPU) rather than a server -- when you turn this on, your browser downloads a real AI model (Meta's Llama 3.2 1B on most devices, Alibaba's smaller Qwen2.5 0.5B on mobile, both distributed by the MLC/WebLLM project) and keeps it cached locally so it isn't re-downloaded on later visits. This is a genuinely large one-time download -- expect it to use several hundred megabytes to roughly a gigabyte of storage and bandwidth, and it needs a browser/device with WebGPU support; if yours doesn't have it, these features stay unavailable rather than failing partway through.
Because generation happens on your own device, the word dumps and messages you send to Benny for this purpose aren't transmitted to us or to a third-party AI provider to produce that reply -- they're still saved to your account the same as any other entry or message, exactly as described elsewhere on this page. This is a deliberate, current choice of models, not a permanent one -- see Settings > About for which model is actually running at any given time.
Student profiles are entered by the parent/guardian who controls the account -- FreeLoom doesn't collect information directly from a child, and children don't have their own login. If you're a parent adding your own child's information, that's what this feature is for. This distinction matters under laws like the U.S. Children's Online Privacy Protection Act (COPPA), but exactly how it applies can depend on specifics we're not positioned to judge for you -- get real legal advice here specifically before treating this section as sufficient.
We use a small number of service providers to run FreeLoom, each only processing what they need to provide their part of the service:
A small number of admin accounts on the FreeLoom team can access account and support data to help with support requests. Any admin request for read-only access to your account requires your explicit approval first, is time-limited, and is visible to you the entire time it's active.
Support message threads auto-delete after a period you can configure (7 days on the Free plan, up to 30 days -- or never, on paid plans). Learning records, transcripts, and portfolios are kept for as long as your account is active, since they're the actual point of the product. If you downgrade to a plan with a lower student limit, existing student profiles beyond that limit are preserved (not deleted) but locked from new edits until your plan covers them again.
The fastest way to reach us is through the Messages feature in your account, if you're signed in. Otherwise, email support@freeloom.io.