Privacy Policy

1. Private by default

Everything you create on Relme — responses, pursuits, profile fields — is private until you explicitly choose otherwise. Making something public is always a deliberate action you take, never a default we set.

2. What we collect

We collect what you give us: your name, email, password (stored only as a secure hash by our authentication provider), and the content you write. We also collect basic technical data needed to run the service — such as IP addresses for security and rate limiting — and a record of the actions you take in your account: for example, that you published a story or approved a suggestion, and the tool calls an app you connected (such as Claude or ChatGPT) makes to Relme, recorded by name. We record that something happened, never the content — never what you wrote, and never what a connected app said.

3. How we use it

To provide Relme: storing your work, rendering your profile to the audiences you choose, sending you the emails you request (confirmation, magic links), and keeping the platform secure. If you use AI features like Build with Relme, your relevant profile content is sent to our AI provider to generate responses — and any suggested change still requires your explicit approval before it touches your profile.

The record of actions in your account helps us understand how Relme is used, so we can improve the product, and lets the Relme team invite feedback from the people using it — a person may write to you to ask what is working. You never have to answer, and we never email you if you have unsubscribed.

4. What we do not do

We do not sell your personal information. We do not show advertising. We do not compute or publish any universal "person score" about you. We do not let agents or third parties silently modify your identity record.

5. Public content

If you make your profile public, the content you marked public is visible to anyone — including through our documented public API, markdown export, and search engines. Switching back to private removes it from those surfaces, but copies already made by others are outside our control.

Vouches. Where vouching is available, another member can write a short statement about work of yours they saw — a vouch — with one line on how they know you. It is their words, about you: it shows on your public pursuit, with their name (even if their own profile is private), only after you approve that exact version, and you can hide it at any time. Its author can edit it (every edit comes back to you for approval; the approved version keeps showing meanwhile) or withdraw it. Every version is kept, readable only by its author and by you. If its author deletes their account, their vouches are removed from every page and from our database; you see a note without their name, for 30 days, saying a vouch left with its author's account. If you ask someone for a vouch by email, we use the address you typed to send that one request — never a reminder — and anything they write before they have an account is held for 7 days and seen by no one but them. Vouches are never counted, scored or ranked anywhere, and we never confirm that what a vouch says is true.

Visits to a public profile. When someone reads a public profile, we count the visit for the profile's owner. Opening the page or one of its public project pages, fetching its markdown version, reading it through our public API, asking its chat a question, or scanning the owner's Relme Card each adds one record. A record holds which profile was read, how, and when — for a project page, which project; for a card scan, whether the code came from a printed card or a wallet pass; whether the reader looked like a person's browser, an AI agent, a web crawler, or another program, judged from the User-Agent string its software sent (and, for automated readers, which one); the name of the site the reader came from (for example, google.com), never the full address; a share-link code, if the link they followed carried one; the source the link was tagged with (its utm_source — for example chatgpt), kept only if it names a site on our fixed list; and a visitor code.

The visitor code is computed from the reader's IP address and User-Agent, the profile being read, a random value we replace every week, and a secret kept only on our servers. It lets us count the same reader once per week on that profile. The same reader gets a different code on every other profile, so codes cannot be matched across profiles, and because the code needs a secret only our servers hold, no one else can compute it. Once a week is over its random value is deleted (at the first visit of the next week); about seven days later, when our database backups have rolled over, that week's codes can no longer be recomputed by anyone, including us. The IP address itself is not stored.

When an automated reader names itself as a known crawler or AI agent, we check the claim the way that company publishes: against its list of IP address ranges, or with a reverse-and-forward DNS lookup of the reader's IP address. We keep only the result — confirmed, failed, or no published method — and a reader that fails is counted as an unverified bot, never under the name it claimed. To avoid repeating the lookup on every read, a server remembers the answer for that address in its memory for up to a day; it is never written to a database. We never run this check on a person's browser.

If the reader is signed in, their account is used only to leave out an owner's visits to their own profile; it is not saved with the record. When owners read their own profile signed in, we also keep that week's visitor code for them, so a visit from the same browser and network that week while signed out — before or after — is recognised as probably theirs: it is left out of their counts and shown to them as a separate number. Those codes are kept only for the current week, no account can read them — the owner's included — and they are deleted with the week's random value. Only the profile's owner is shown these counts, as totals — how many reads, of what kind, from which sites, sources and share links — never as a list of individual visits, and they never appear on a public page or through our public API. On a public profile page, part of the count is sent from the reader's browser: it stores nothing on the reader's device, and the analytics choice on our Cookies & browser storage page does not switch it off.

Visit records are kept until we delete them. That is a manual step today, not an automatic one, and we plan to remove records once they are about 400 days old. All of a profile's visit records are deleted when its owner deletes their account.

6. Service providers

We rely on a small set of processors to run Relme: Supabase (database and authentication), Vercel (hosting and analytics), and AI model providers for the features that use them. Each receives only what it needs to do its job.

7. Your rights and controls

You can view, edit, export, and delete your content at any time from your account. You can request a copy of your data or full account deletion by emailing us. Depending on where you live, you may have additional statutory rights (such as under GDPR or CCPA), and we intend to honor them.

8. Security and retention

Access to your data is protected by authentication and row-level security in our database. We keep your content for as long as your account exists; when you delete it, we remove it from active systems within a reasonable period, subject to legal retention requirements.

9. Changes and contact

We will update this policy as the product evolves and give notice of material changes. Questions or requests: hello@relme.io.