beeperbox

subs · ·

← home · //product-release

0

beeperbox

scrawny-crawdad · //product-release ai · 14d ago · 0 replies

I wanted my AI agents to respond to customers on whatever channel they came from — WhatsApp, Signal, iMessage, Telegram, Instagram, Messenger, whatever — with one unified chat history across all of it.

Customers reach out where they already live. Some on WhatsApp, some on Telegram, some in Instagram DMs. If my assistant only lives on one platform, I either lose leads or force people to switch apps. And I wanted one continuous history so context carries when a conversation moves from WhatsApp today to Signal next week.

First stop was OpenClaw 50+ messenger connectors in one project. But "50 connections" means 50 things to maintain. Most of those connectors are reverse-engineered or community-kept, they break on platform updates, and I'd be spending weekends fixing bridges instead of running a business.

So I flipped the problem. Instead of 50 connections, one. Beeper already bridges WhatsApp, iMessage, Signal, Telegram, Discord, Slack, Messenger, Instagram, LinkedIn, Matrix, and more — and they maintain the bridges as their product. I don't touch a single connector. My chatbot talks to Beeper, Beeper talks to every network.

On top of that I built multis — the personal/business assistant chatbot — against Beeper Desktop's HTTP API. It reads incoming chats across all channels, keeps one unified history, and lets AI agents respond on whichever channel the customer originally used.

Then the obvious next problem: Beeper Desktop is a GUI Electron app, not exactly server-friendly. I wanted multis alive 24/7 on my home server, not tethered to a laptop. So I wrapped it in Docker — virtual display, one-time browser login through noVNC, and an opinionated MCP server on top so Claude Code, Cursor, Cline, bareagent, or any MCP runtime can plug in through one standard protocol.

That's beeperbox — the Docker piece I extracted from multis so others can use it independently. Messenger backbone in one container, multis (or your own agent) is the brain on top. Multi-arch, runs on Raspberry Pi, Oracle free ARM, Apple Silicon, or any x86 VPS.

Use it if: you want AI agents reaching customers across many networks with unified history, without babysitting 50 connectors.

Don't use it if: you only need Telegram. BotFather library, 5 lines, done.
Free, MIT, self-hosted.

Anyone else running a multi-channel agent setup? Curious how you solved the "one brain, many channels" problem.

https://github.com/hamr0/be...github.com

// comments · sort:

bestnew

no comments yet — be the first.

0 / 10000