PRIVACY POLICY
Your memory, with receipts.
Myeline is a memory layer for AI assistants, operated by Sandbox Group LLC (“we”, “us”). It gives Claude, ChatGPT, and other MCP-compatible clients a memory you control. Every fact has a source attached, nothing an AI proposes goes live until you approve it, and when you remove a fact you have to give a reason. This policy covers myeline.si, workspace.myeline.si, and the hosted cortex at cortex.myeline.si (together, “the Service”).
The short version
- We store the facts you and your connected AI clients save to your cortex, the sources attached to them, and the audit trail of what happened to each one.
- AI clients can only propose memories. Proposals wait in your review queue and do nothing until you approve them.
- We refuse to store Social Security numbers, bank account numbers, payment card numbers, and credentials such as passwords and API keys. Writes that contain them are rejected before anything is saved, including to the review queue.
- We don’t sell your data, use your cortex content for advertising, or use it to train AI models. Our public pages and sign-up page use a Google Ads tag to measure and target our own ads; it never runs inside the workspace.
- A small number of vendors, listed below, help us run the Service.
Information we collect
Account information. When you sign up, our authentication provider (Clerk) collects your email address, your name if you give one, and sign-in details. We keep a link between your Clerk identity and your Myeline workspace.
Your cortex content. This includes:
- facts and memories saved through the workspace or through connected AI clients
- the provenance for each fact: where it came from, which client proposed it, and when
- any tags you add
- review decisions (approve or reject), retractions, and the reasons you give for them
- sensitivity flags our PII gate adds to staged facts
- search text you send, whether you search in the workspace or through a connected AI client
Connected-client and OAuth data. When you connect an AI client, we store the OAuth client registration, short-lived authorization codes (5 minutes), and access tokens (30 days by default). We store each token’s scopes (query, propose, retract) and a label naming the client.
Billing information. If you subscribe, payment happens on PayPal. We never see or store your card or bank details. We keep your PayPal subscription ID, your plan status, your paid-through date, and a log of billing events.
Technical data. Our servers keep operational logs (request handling, errors, health checks) to keep the Service running and secure. The workspace saves one setting in your browser’s local storage, which remembers whether you closed the setup card. Clerk sets the cookies needed to keep you signed in. On our public pages (the home page, the docs, and this policy) and the sign-up page, the Google Ads tag (gtag.js) sets cookies that tell us whether a visit came from one of our ads and whether it led to a sign-up, and lets Google show our ads to people who have visited these pages. Google receives the page address, ad-click identifiers, and browser and device data. The tag never loads inside the signed-in workspace, so it never sees your cortex content, searches, or tags. We don’t use third-party analytics cookies.
How we use information
- to provide the Service: store your cortex, answer queries with sources, and run the review queue
- to authenticate you and the AI clients you connect
- to create search embeddings so your cortex can find related facts, not just exact word matches
- to enforce plan limits (the free plan allows 200 live facts) and process subscriptions
- to secure the Service, prevent abuse, and fix problems
- to contact you about your account or changes to this policy
Sensitive information safeguards
Every write passes through a PII sensitivity gate before it is saved.
- Hard floor (always on): We reject Social Security numbers, bank account and routing numbers, payment card numbers, and credential-like strings. We never store them, not even in the review queue.
- Comfort tiers you choose: open, personal (the default), or strict. These decide which other details, like addresses, phone numbers, emails, birthdays, names in sensitive contexts, and employer or client mentions, get flagged on the review card so you can check them before approving.
How AI clients use your data
When you connect an AI client such as Claude or ChatGPT, that client sends queries and proposed memories to your cortex, and gets back answers with their sources. Once an answer reaches the client, that provider’s own privacy policy applies to it. You choose which clients to connect, and you can disconnect any of them at any time in Settings, which revokes their access immediately. Connected AI clients are not our subprocessors.
Subprocessors
We use these vendors to process personal data on our behalf:
| Subprocessor | Purpose | Data processed | Location |
|---|---|---|---|
| Clerk, Inc. | Authentication: sign-in, sign-up, sessions | Email, name, sign-in and session data | United States |
| Neon (Databricks, Inc.) | Managed Postgres database for all cortex, OAuth, and billing records | All account and cortex content described above | United States — AWS us-west-2 (Oregon) |
| DigitalOcean, LLC | Application hosting (the workspace and the cortex service) | All data in transit through the Service; operational logs | United States — San Francisco (sfo3) |
| OpenAI, L.L.C. | Text embeddings for semantic search (text-embedding-3-small) | Fact text and search text, sent to create embeddings | United States |
| Google LLC | Advertising measurement and remarketing (the Google Ads tag) on the public pages and the sign-up page | Pages visited, ad-click identifiers, cookie IDs, browser and device data, IP address; never cortex content | United States |
| PayPal, Inc. | Subscription payments and billing verification | Subscription ID and status; PayPal handles payment details directly | United States |
Under OpenAI’s API data-use terms, OpenAI does not use API inputs to train its models. If semantic search isn’t available, we fall back to a local word-matching method that sends nothing to OpenAI. We will update this list before adding a new subprocessor that handles your cortex content.
Sharing
We share personal data only with the subprocessors above, with AI clients you connect, when the law requires it, to protect the Service and its users, or as part of a merger or acquisition (with notice to you). We do not sell or rent personal data.
Retention and deletion
- Retracted facts stop appearing in answers and lists. We keep a record of the retraction and its reason in your audit log, so your memory history stays complete.
- Rejected proposals are never published, and are kept only as part of your review history.
- OAuth codes and tokens expire automatically (5 minutes and 30 days). Disconnecting an app in Settings revokes them immediately.
- Account deletion: Email hello@myeline.si from the address on your account and we will delete your workspace, including facts, sources, audit logs, and tokens, within 30 days. The only exceptions are records the law requires us to keep, such as billing records.
- Backups are overwritten on our database provider’s normal backup schedule.
Security
Every request runs under a database role with only the permissions it needs, and each tenant’s data is isolated. Connections are encrypted with TLS. Tokens are scoped, and no OAuth scope can approve a fact or change billing; only you can, signed in. PayPal webhooks are accepted only after their signatures are verified. No system is perfectly secure, but we built Myeline so every change to your memory leaves an auditable record.
Your rights
Depending on where you live, including the EU, UK, and US states such as California and Oregon, you may have the right to access, correct, delete, or export your data, or to object to or limit how it’s processed. You can see and retract your facts directly in the workspace. For anything else, email hello@myeline.si. We won’t treat you differently for using these rights.
International transfers
We are based in the United States, and our subprocessors process data mainly in the United States. Where the law requires it, we rely on appropriate safeguards such as Standard Contractual Clauses.
Children
The Service is not for anyone under 16, and we do not knowingly collect data from children.
Changes
If we make material changes, we’ll post the updated policy here and let account holders know by email before the changes take effect.
Contact
Sandbox Group LLC, Portland, Oregon, USA. Email hello@myeline.si.