How does PriChat work?

PriChat is an end-to-end encrypted messenger. That means only you and the person you talk to can read your messages, no one else, not even us as the operator. You don\'t need a phone number, just a username of your choice, and you can start right in the browser or as an installed app.

Sounds like everyone else at first. The difference is in how it is built. Let us look at it step by step.

When you sign in, a strong key (AES-256) is computed from your password right on your device. This uses a deliberately slow procedure (PBKDF2, 600,000 rounds) to make password guessing expensive. This key lives only in memory and is never stored.

For logging in to the server, a second value is derived separately, with a different salt. Only this login token goes to the server, and only as an argon2id hash. Your password and your data key stay with you. That is why the server technically cannot decrypt your local data at all.

PriChat uses the same proven approach as Signal: an X3DH handshake to establish a shared secret, and then the Double Ratchet, which generates a fresh key for every single message. This is called forward secrecy: even if one key is ever exposed, past messages stay protected.

The devices exchange public keys via the server, never private keys. Only if your counterpart hasn\'t published modern keys yet does a simpler fallback kick in. Group chats, by the way, are not a separate technology: the same message is encrypted individually for each member, and the server never learns that it belongs to a group.

This is the most honest part. A server needs to know a minimum to deliver messages. But content is not part of that. On the left is what actually lives on the server, on the right what provably never arrives there.

One point matters a lot to us here, and it is new: your social circle stays private too. Your contact list only sits on the server as an encrypted bundle, and even the sender of a friend request is encrypted anonymously against the recipient's key. So the database holds no plaintext link “X wants to be friends with Y”.

And so you can see this is not a marketing claim: this is what the data in the database actually looks like. An undelivered message sits only as a ciphertext block in the content field (recognizable by the DR: or ENC: prefix), never as readable text. Your password is stored as an argon2id hash, your contacts as an encrypted bundle.

You'll surely notice the empty from column, and that is deliberate. Who sent a message appears nowhere in plaintext: the sender travels encrypted inside the content field and can only be recovered by the recipient. The server sees only which mailbox (to) something was placed in, so it can deliver it, never who it came from.

Delivered messages are removed from this table immediately, or after 14 days at the latest if your device was offline for a long time. So the server keeps no message history. That matters for the next point.

Encryption protects your messages in transit. But there is a second front: your device and your account itself. For that you can add several locks on top of your password, all optional, all switchable right in the settings.

Two-factor (2FA). You set up an authenticator app (via QR code, plain standard TOTP). With 2FA on, the server additionally asks for the 6-digit code at sign-in. Even someone who knows your password cannot get in without that second device.

Passkey and biometrics. Instead of typing your password every time, you unlock the app with Face ID, fingerprint or your device PIN. There is more to it than it looks: your data key is not simply stored, it is wrapped and encrypted with a secret from your device's security chip. That secret never leaves the chip. The server only ever sees an opaque block, so again there is no spare key on our side.

App lock. With one tap you lock the app. Your key is immediately cleared from memory, and after that nothing is readable without your password, passkey or recovery phrase. If you like, the app also locks by itself, after 5, 15 or 30 minutes of inactivity.

To be clear: 2FA and passkey protect access to your account and your device. The end-to-end encryption of your messages runs independently of that, always.

Encryption is only as good as your certainty that you are really talking to the right person, and not to someone who slipped in between. For exactly that there are two visible checks.

Safety numbers. For every contact there is a number derived from both your keys. It is identical on both devices as long as no one intervenes. Compare it once (read it out, send a photo, however you like) and you have certainty. Verified contacts are permanently marked with a checkmark.

Key-change warning. The app automatically remembers a contact's first key. If a different one shows up later, for instance because someone tries to interpose themselves, you get a warning instead of quietly continuing. This protects you even with contacts whose number you never explicitly compared.

New-login warning. If an unknown device signs in to your account while a trusted one is active, a warning appears there with IP, device and time. You can then refuse to approve the new device.

Sometimes it is not about attackers on the network, but about the situation right in front of you: a phone being checked, a forced handover, a lost device. For those cases there is the emergency wipe.

One move wipes the local messages and keys of the contacts you selected in advance, right on your device, and signs you out. In the settings you decide which contacts are affected, whether their contact entries disappear too, and whether it asks once more first or happens immediately.

Say you registered on your phone and later sign in on your laptop. What happens then? In short: you simply sign in with username and password, and your key is freshly computed on the new device.

One more detail: for security reasons only one device is active at a time. If you sign in on the laptop while the phone is still online, you are asked to sign out there first.

When you sign up you get a phrase of 12 words. Keep it safe, because it is your emergency key. With it you can restore your messaging identity and your keys and set a new password. The words never leave your browser, the server does not know them.

Honest here too: the phrase brings back your keys and your identity, not your old message history. That history simply does not exist on the server, so no recovery can conjure it up.

No system is perfect, and I would rather tell you openly where the edges are:

• The simple fallback (when the other side has no modern keys) and call signaling do not have forward secrecy. For two up-to-date devices the modern Double Ratchet always runs.
• Support requests are intentionally not end-to-end encrypted, because they go to the operator, like at a normal help desk.
• Profile and group pictures live as metadata on the server, not end-to-end. If you prefer, leave them empty.