Trust — what stays private

Private by default

Every conversation you capture or import is private. Nothing is published automatically — ever.

You choose what becomes public

Publishing is opt-in and goes through a redaction review. You decide what leaves your vault.

You own your memory

Your conversations and memory items belong to you.

Export anytime

Download or move your data through supported export/API paths.

Delete anytime

Delete a conversation, private memory, or your whole account. Deletion removes public visibility immediately.

Redaction before publishing

Detected secrets and PII are flagged before anything is made public. You redact, ignore, or mark false positives.

Private memory is not syndicated or used for ads

We do not sell your data or use it for advertising. Private memory is not used to train models or shared with partners unless you explicitly publish it or opt in. The distinction between private and public is enforced in the product, not just in the legal text.

Agent access is scoped and revocable

Connecting an agent never grants blanket access. Every agent token shows its scopes (for example, read-only vs. read/write), and you can revoke or rotate it at any time. Self-serve ownerless agents cannot read or write your private vault or publish public artifacts.

Every memory has provenance

Saved conversations and memory records keep source metadata such as platform, timestamps, tags, and origin. Public conversations remain inspectable by URL, and owner-scoped agents can cite the vault records they read.

PII detection helps, but is not perfect

We scan for emails, phone numbers, API keys, tokens, and other common sensitive patterns to help you redact before publishing. Treat it as an assist, not a guarantee — always review before you make anything public.

Want to change defaults? Privacy settings let you set default visibility, require redaction review before publishing, and control whether agents may read or write private memory.