Overlap
Privacy, plainly
Overlap only works if you can be honest with it — so trust isn’t a policy page here, it’s the product. No summary, no legalese: here is exactly what happens to what you say.
Introductions are double opt-in
You are never revealed to anyone unless both people say yes. Sensitive topics (health, grief, relationships, money) are never matched by default — you’d have to turn them on yourself. You can decline or say “never for this” at any time, and nobody is told that you passed.
What we store
Your account (email), your conversations with the assistant, and — for any past conversations you choose to add — only a derived topic and a short summary. That applies to Claude Code and Codex session files and to a ChatGPT or Claude history export alike: the file is opened and previewed in your browser, we never receive the raw transcript, and only the conversations you tick are sent to us.
What the assistant remembers
To be more useful over time the assistant keeps a few short notes (ongoing projects, stable preferences) — never raw conversations. Every note is visible on the Memory page, where you can delete any of them or switch memory off entirely.
A human makes the introductions
In this early phase the matches aren’t made by an algorithm — a curator proposes introductions by hand, which is why they’re rare and why they’re good. It also means a person may read what you write. We’d rather tell you that plainly at sign-up than bury it here.
What the curator actually sees
Their screen doesn’t open on your conversations. Each one appears as the topic and one-line summary it was distilled to, with whether it’s eligible for matching — that’s what an introduction is proposed from. Reading the actual messages is a separate, deliberate click, and every time someone takes it we write a line to your record saying so. You can ask us for that list.
Keeping a conversation out of it
Any conversation can be taken off the record from the button under the message box. Off-record chats aren’t distilled into a topic, aren’t matched, and don’t appear on the curator’s screen at all — not even as a locked row. Switching it on also stops matching on what that conversation already taught Overlap.
Your controls
You can take any conversation off the record, see and manage the topics you’re matchable on, and leave any introduction. On your profile page you can download everything we hold about you as a file, or delete your account — we email you a link to confirm, and then your conversations, topics and introductions go. If you’d rather a human did it, email [email protected].
How long we keep things
Conversations and topics stay until you delete them, or until you delete your account. Files you attach to a chat are different: we drop them after 90 days. Deleting a message or a conversation takes its attachments with it immediately.
What we keep after you leave
When you delete your account we keep one row, and it holds no readable detail about you: a one-way fingerprint of your email address, the date, and a count of how many rows were removed. It exists so that if you ever ask “did you really delete my account?”, we can answer. It cannot be turned back into your address, your name, or anything you wrote.
What we don’t do
We don’t sell your data, we don’t read your ChatGPT/Claude web history (we can’t — you bring only what you choose), and we don’t match sensitive topics without your explicit opt-in.