Docs / Security & Reliability

Security & Reliability

Your prompts, rules, and docs are your working knowledge. We treat them that way.

Security

Encrypted everywhere. All traffic is HTTPS/TLS and data is encrypted at rest. Plain HTTP is redirected before it carries a byte.
Passwords we can't read. Stored only as bcrypt hashes (12+ character policy). Three failed sign-ins lock the account until it's recovered.
Sign in with Google or Microsoft (optional). Use either on the sign-in and sign-up pages, or connect one from Settings → Connected accounts as an additional way in. We never attach a provider to an existing account automatically — you sign in and connect it yourself — and you can't disconnect your only remaining way to sign in.
Two-factor authentication (optional). Add an authenticator-app code on top of your sign-in — password or Google/Microsoft alike — with one-time backup codes. With it on, a stolen password isn't enough.
API tokens as hashes. We store a SHA-256 hash, never the token itself — it's shown to you once. Revoking a token cuts off access instantly, everywhere it was used.
Sessions done right. httpOnly, secure, same-site cookies. Changing or resetting your password signs out every other device on the spot.
You see every sign-in. A devices list shows every active session (sign any out), and we email you when your account signs in from a new device.
Company controls. Roles on every workspace, VPN/IP allowlisting, instant member revocation, and a company audit log.
Payments never touch us. Billing runs on Stripe; your card number goes to Stripe's vault, not our servers.

Reliability

Every change is undoable. Full version history on every item — restore any previous version in one click. Anything you delete waits in the Trash for 30 days.
No silent overwrites. Optimistic locking catches two people editing the same item and warns instead of clobbering; you can see who else is in an item.
Backed up automatically. The whole database is backed up and encrypted automatically — a disaster-recovery net so a bad day for our servers isn't a bad day for your library.
Grace, not cliffs. Go over a limit or let a plan lapse and libraries turn read-only with a 90-day grace window and email warnings — nothing vanishes out from under you.
Idle never means deleted. We don't purge accounts for inactivity. Step away for a year — your library is waiting.

Your data & privacy

What you create is yours. Your personal library is yours; a team library belongs to that team; a company library belongs to that company. Nothing is shared between accounts. Another customer — including one who signs up and pays — can't see your items, search them, or reach them by guessing an address: access comes from your membership in a library, and every read is checked against it. Two things cross accounts, both deliberate and both visible: the prebuilt AIStorageDepot library (ours, read-only, everyone can browse it) and a share link you create yourself.
Never used to train models. Your library is your private working knowledge. We don't sell it, mine it, or train anything on it. The only AI that sees your prompts is the one you send them to — there's no AI inside AIStorageDepot at all: "Open in Claude" hands the prompt from your browser to your AI account.
The honest limits. Encryption at rest protects the disks and backups; it isn't zero-knowledge, so your prompt content is readable to us — the same trade any service that searches your content for you makes. Isolation between accounts is enforced by the application over a shared database rather than by the database itself, so the guarantee is our code and the tests that pin it. We hold no SOC 2 or ISO 27001 and haven't commissioned a third-party penetration test. If that matters for your evaluation, ask us rather than assume. Organizations that can't put prompts in shared SaaS at all can run the self-hosted edition — your own Azure or servers, your own database, your content never leaves your network.
No lock-in. Download any item as a real file, export a .zip, or pull everything over the API and MCP. You can permanently delete your account yourself from Settings — no support ticket.