Relay is a personal AI agent that replies to your WhatsApp messages when you can't. It reads the conversation, follows rules you set, and stays fully under your control from a private dashboard.
Every reply is generated from real conversation context and the rules you've set — not a canned auto-responder.
Reads recent conversation history before answering, so replies feel natural instead of context-blind.
Enable, block, or add custom instructions per contact — all editable live from the dashboard, no restarts.
Stays quiet in groups until it's tagged, then reads the last several messages from everyone before replying.
Runs on your own server behind a password + authenticator login. No third party ever touches your messages.
No app to install on the other end — Relay pairs the same way WhatsApp Web does.
Link your WhatsApp number with a QR code from the dashboard's Status tab.
Choose who gets auto-replies, add custom instructions per contact, and schedule away mode.
Incoming messages are checked against your rules, then answered in a short, natural tone.
Watch the conversation log live, and flip the global kill switch anytime.
No black box — trace exactly how an incoming message becomes (or doesn't become) an outgoing reply.
Most auto-responders use templates. Relay builds a private behavioral model from real conversations — capturing how you actually write, not how you should write.
Behavioral signature · unique to each user · never leaves your server
Passively observes real messages to isolate what makes your writing style uniquely yours — across a dozen behavioral dimensions.
For each person you talk to, the model reads their communication style and shifts output register to feel like a genuine two-way conversation.
The behavioral signature is injected at the inference layer — not the prompt surface. Style stays locked even as topics change.
No shared infrastructure, no vendor lock-in — the whole agent runs where you tell it to.
Deploy it on your own server in a handful of commands.