Help guides
Data Policy
How Memory Currents stores and processes content, what export includes, and what deletion can leave behind.
Last updated: October 26, 2026
How we handle your content
We use your stored content to operate and support Memory Currents. We use it for product improvement only with your explicit consent. We may use aggregate usage information to improve the product.
Saving content or accepting the general agreement is not permission to keep that content for later product evaluations. That use needs separate permission covering the content and purpose. Investigating a reported problem is support.
We do not sell your memories or other content, or use them for advertising. The Privacy Notice names the operator, describes the information collected and recipients, and explains how to make a privacy request.
Storage and operator access
Source captures are stored in Amazon Web Services object storage. Neon runs the database holding Memories and the information organized around them, along with account information, preferences, operational records, and feedback. Source captures can have multiple stored versions. Database recovery history and processing or delivery records are separate from the information visible in your library.
Authorized operators can access content when needed to operate and support Memory Currents. Access is not technically limited to content you separately approve for a support read, and we do not promise a per-read permission or access-audit system.
AI processing and connected applications
Memory Currents sends content to Anthropic to power features such as splitting Memories into Thoughts, describing content, and organizing related information. Requests include a consistent, scrambled account identifier for abuse detection. Because it stays the same for your account, it is not anonymous.
Google Gemini receives Memory, Thought, and Stream text to create embeddings. These are numerical representations of meaning that help Memory Currents recognize related information. The embeddings are stored in the account's database.
AI processing sends content outside Memory Currents' own storage. Provider retention is separate from changes to your library or removal of account records; removing information here is not a promise of immediate removal from a provider's records. Processing may take place outside the United States.
If you turn on AI connections and authorize an application such as Claude or ChatGPT, that application can receive information through the tools you approve. Its own terms and data practices apply. Stytch handles the connection's authorization and account information; our service sends tool results to the connected application. The AI connections guide explains what connected tools can do.
Retention and deletion
We keep active content and account information for the account's use, subject to the limits of the available removal processes. Different kinds of records have different lifetimes. We do not promise one retention period or deletion deadline for every copy.
Memories and account requests
You can edit or archive a Memory. Archiving sets it aside and is reversible; it does not erase the source capture, earlier versions, or every related record.
There is no in-app account-deletion control. To request account deletion, email support@memorycurrents.com. Our current process cannot completely remove a live account and its data from every system. A request is not confirmation that removal has happened or that every copy is gone.
Records used to route your sign-ins and link your contact channels remain unless separately removed. Removing account content does not by itself remove those records.
Recovery copies and source versions
Database recovery history and stored source versions are distinct from the current records you see. Their existence does not provide a complete backup you can restore yourself.
Processing, delivery, and operational records
Account removal does not sweep every record used to run the service:
- Recent processing history can contain your words, including text produced while working with your Memories. It can remain after account deletion until the service holding that history deletes it.
- Email duplicate-prevention records can retain an email address until their storage service deletes them. The time during which a record prevents duplicate processing is different from the time when it is physically removed.
- Temporary delivery and retry records can contain messages, sender details, or record identifiers while a delivery is being handled or retried. Records left by an interrupted delivery can remain until storage cleanup.
- Failed-processing records can contain content and account information from a delivery, import, or other operation that failed. They remain available for investigating and recovering from failures until the storage service removes them.
- Abuse-prevention counters can retain account-related identifiers or information derived from network addresses. A counter ceasing to count does not mean its stored record has already been deleted.
- Operational logs support operation, troubleshooting, and security. They can contain account identifiers, email or network addresses, and occasionally fragments of submitted content in database errors. They remain outside the account-removal process until their log retention removes them. Browser diagnostics and alerts also have their own retention.
These records are for operation, support, and security. Content in them is still content; keeping it there does not give permission to reuse it for product improvement.
Feedback
By sending feedback, you agree that we may use it to improve Memory Currents. This permission covers the feedback you submit, not your other stored content. Feedback may remain after account deletion until the operator removes it. Removing its account link does not anonymize identifying information you typed into it. Contact support if you want to request its removal.
Exporting a copy
Export downloads a readable YAML file of the Memories and Currents in your chosen scope. You can include archived records. For Re-Import carries their text, dates, statuses, and Current links. Custom can also carry titles, types, and Thoughts with their current To-Do statuses. Neither mode carries tags. Edited Memories export their current text, not earlier versions.
The file is not an export of every stored record or a complete account backup. It does not include Streams, histories, source captures, attachments, phone numbers, or preferences. Import ignores the extra titles, types, and Thoughts in a Custom file; it does not restore those Thoughts or their To-Do statuses. An unsuccessful export does not mean you received a complete copy; follow the message shown in the app and contact support if needed.
Health information
Memory Currents is not HIPAA compliant. You decide whether to enter personal health information. We do not provide special handling for health information; the storage, AI processing, retention, and deletion practices described here apply to it too.
Questions and updates
Memory Currents operates the service. Email support@memorycurrents.com with questions or data requests. See the Privacy Notice for applicable rights and the request path.
We update this policy at this address and show its last-updated date above. We do not promise to email every change; applicable notice and consent requirements still apply.